Fstat 294

FSTAT (version 2.9.
4), a program (for Windows 95

and above) to estimate and test population genetics
parameters.
Jérôme Goudet
Department of Ecology & Evolution,
BB, Lausanne University,
CH-1015 Dorigny
Switzerland.
E-mail: jerome.goudet@ie-zea.unil.ch
1
http://www.unil.ch/izea/softwares/fstat.html
21st November 2003
1
This software could not have been developed without the financial support of the
Swiss National Science Foundation, Grants N◦ 31-43443.95, 31-55945.98 and 31- .2000. I
am indebted to the many users who sent me comments, bugs or cheerful messages that
helped improve Fstat: François Balloux, Lars Berg, Michel Chapuisat, Thierry De Meeüs,
Loı̈c Degen, Greg Douhan, Guillaume Evanno, Arnaud Estoup, Laurent Excoffier, Pierre
Fontanillas, Luca Fumagalli, Barbara Giles, Chris Gliddon, Alexandre Hirzel, Michael
Krawczak, Martin Lascoux, Lance Barrett-Lennard, Margaret Mackinnon, Patrick Meir-
mans, Eric Petit, Alan Raybould, Michel Raymond, Max Reuter, François Rousset, San-
drine Trouvé and many others, whose name slipped my mind.
Contents
1 Introduction 4
1.1 What the program FSTAT does . . . . . . . . . . . . . . . . . . . . . 5
1.1.1 What’s new in version 2.9.4? . . . . . . . . . . . . . . . . . . 6
1.2 What the program FSTAT does NOT do . . . . . . . . . . . . . . . 7
2 Running the program 8

2.1 Creating an input file . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2 LIMITS IN FSTAT . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3 File menu 10
3.1 Open submenu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.2 Save Default Options subMenu . . . . . . . . . . . . . . . . . . . . . 10
3.3 Exit subMenu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
4 Choose options menu 11

4.1 Label file for pops sub-menu . . . . . . . . . . . . . . . . . . . . . . . 11
4.2 Loci to use sub-menu . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
4.3 Samples to use sub-menu . . . . . . . . . . . . . . . . . . . . . . . . 11
5 Per locus and sample statistics 13

5.1 Allele frequencies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
5.2 Genotypic frequencies . . . . . . . . . . . . . . . . . . . . . . . . . . 13
5.3 Number of alleles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
5.4 Allelic Richness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
5.5 Gene diversities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
5.6 Fis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
6 Global statistics 15
6.1 Nei’s estimators of F-statistics . . . . . . . . . . . . . . . . . . . . . . 15
6.2 Weir & Cockerham estimators of F-statistics . . . . . . . . . . . . . 16
6.2.1 Jackknife variance and Bootstrap confidence interval . . . . . 17
6.3 RST . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
6.4 FST per pair . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
7 Testing 19
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
7.2 Global tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
7.2.1 Hardy Weinberg Within samples . . . . . . . . . . . . . . . . 21
7.2.2 Hardy Weinberg Overall samples . . . . . . . . . . . . . . . . 21
7.2.3 Population Differentiation NOT assuming Hardy Weinberg
within . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
7.2.4 Population Differentiation assuming Hardy Weinberg within . 21
7.3 Tests per sample or pair of samples . . . . . . . . . . . . . . . . . . . 21
1
CONTENTS 2
7.3.1 HW tests per locus and sample . . . . . . . . . . . . . . . . . 21

7.3.2 Pairwise tests of differentiation . . . . . . . . . . . . . . . . . 21
8 Composite disequilibrium 23
8.1 Estimating composite disequilibrium . . . . . . . . . . . . . . . . . . 23
8.2 Testing for genotypic disequilibrium . . . . . . . . . . . . . . . . . . 25
9 Run button 27
10 Comp. Among groups of samples 28

10.1 Principle of the tests . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
10.1.1 Number of groups . . . . . . . . . . . . . . . . . . . . . . . . 29
10.1.2 Define groups . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
10.1.3 Type of tests for two groups . . . . . . . . . . . . . . . . . . . 29
10.1.4 Statistics to compare among groups . . . . . . . . . . . . . . 29
10.1.5 Number of permutations . . . . . . . . . . . . . . . . . . . . . 30
10.1.6 Run . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
11 Biased dispersal 31
11.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
11.2 Design of the tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
11.2.1 Assignment index AI . . . . . . . . . . . . . . . . . . . . . . 31
11.2.2 FST . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
11.2.3 FIS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
11.2.4 HO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
11.2.5 HS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
11.2.6 Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
11.3 Running the program . . . . . . . . . . . . . . . . . . . . . . . . . . 33
11.3.1 Input file format . . . . . . . . . . . . . . . . . . . . . . . . . 33
11.3.2 Running the program . . . . . . . . . . . . . . . . . . . . . . 34
11.3.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
12 Mantelize it! 36
12.1 Matrices to Columns . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
12.2 Multiple regression and partial mantel . . . . . . . . . . . . . . . . . 36
13 Utilities menus 38
13.1 Reset seeds Random number generator . . . . . . . . . . . . . . . . . 38
13.2 File conversion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
14 Files of results 39
15 Special use of Fstat 41

15.1 Haploid data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
15.2 Only one sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
15.3 Pooling samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
15.4 Missing data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
A Examples of input files 43

A.1 One digit encoding for input data file . . . . . . . . . . . . . . . . . . 43
A.2 Two digits encoding for input data file . . . . . . . . . . . . . . . . . 43
A.3 Example of label file . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
A.4 Example of file format for testing biased dispersal . . . . . . . . . . . 44
A.5 Example of input file for multiple regression and partial mantel . . . 44
CONTENTS 3
B Files of results from the input file in Appendix 1. 45

B.1 DIPLOID.OUT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
B.2 Files obtained from the ”FST per pair” option . . . . . . . . . . . . . 47
B.3 DIPLOID.X2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
B.4 Files obtained from the ”testing pairwise pop differentiation” option 48
B.5 Example of result file from the comp. Among group of samples tab
sheet file diploid-test.out of the distribution . . . . . . . . . . . . . . 49
B.6 Example of result file from the biased dispersal menu . . . . . . . . . 49
B.7 Example of result file from the multiple regression and partial mantel
menu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
Chapter 1
Introduction
For evolution to occur, there must be polymorphism on which selection can

act. The amount of polymorphism at a locus results from the interaction of the
following forces: selection, genetic drift, mutation and migration. By screening
genetic markers on their favorite species, population biologists attempt to estimate
the strength of each of these forces. The quantities that they are likely to find useful
are the allele frequencies and the genetic diversity of the markers, the proportion
of each genotype and their expected proportion under the Hardy Weinberg null
model. Usually, they will have samples from many locations and might want to
know whether these locations differ in allele frequency.
Fstat is a user-friendly program that estimates and tests all these statistics.
Its main strengths compared to other softwares are its ease of use and its exhaus-
tiveness.
I’ve paid particular attention to the writing of the help file (this document),
which will hopefully provide guidelines to a newcomer in the field of population
structure analysis.
4
CHAPTER 1. INTRODUCTION 5
Do not hesitate to email me if you have suggestions for improvements, discover

bugs, or simply think Fstat is great!!!
Hardware requirements
A PC running Windows 95 or above. It might work with windows 3.1 and wins32,
but I have not tried it. . . I guess a Pentium processor would be a necessity if ran-
domisations are to be carried out. You should also be able to run it on a (recent
and powerful) Macintosh with PC emulation software.
Cite as
Goudet, J. 2003. Fstat (ver. 2.9.4), a program to estimate and test population ge-
netics parameters. Available from http://www.unil.ch/izea/softwares/fstat.
html Updated from Goudet [1995]
1.1 What the program FSTAT does

From a data set of codominant genetic markers (see input file format), Fstat
calculates the following:
• Number of individuals per sample and loci.
• Allele frequency estimated per sample and overall.
• Observed and expected number of each genotype per sample and locus.
• Unbiased gene diversity per sample and locus.
• Number of alleles sampled per locus and sample, as well as overall.
Fstat Allelic richness per locus and sample, as well as overall samples.
• FIS per locus and sample, as well as a test of whether it is significantly positive
or negative (significant deficit and excess of heterozygotes respectively).
• Nei [1987] estimators of gene diversities and differentiation.
• Weir and Cockerham [1984] F (FIT ), θ (FST ) and f (FIS ) estimated per
allele, per locus and overall. Fstat also calculates Hamilton [1971] relatedness
[relat=2FST /(1+FIT )], calculated using an estimator strictly equivalent to
Queller and Goodnight [1989]. This measure is the average relatedness of
individuals within samples when compared to the whole data set.
• Confidence intervals based on resampling schemes are provided for Weir and
Cockerham [1984] statistics:
– Jackknifing per locus over samples is performed. The resampling unit
are the different samples. This procedure is only carried out if there are
more than 4 samples.
– Jackknifing over loci. In this case, the resampling units are the different
loci. This procedure is only carried out if there are more than 4 loci.
– Bootstrapping over loci. Bootstrapping over loci is performed only when
there is more than 4 loci in the data set.
• Estimation of R-statistics (Slatkin [1995]), specifically designed for microsatel-
lite undergoing stepwise mutations.
• Estimation of FST (θ) per pair of samples.

• Overall test whether samples at each locus are in HW equilibrium.
• Test whether the entire data set is in HW equilibrium.
• Test whether samples are differentiated assuming either that there is HW
within samples or that there is not (Only one of these 2 tests can be carried
out).
• Test whether each sample at each locus is in Hardy-Weinberg (HW) equilib-
rium.
• Test whether each pair of samples is differentiated. The tests do not assume
random mating within samples. A table of significant pairwise differentiation
after corrections for multiple testing is produced.
NEW, Fstat Estimates two multi allelic coefficients of composite disequilibrium, R and ∆0 .
These coefficients are inspired from Weir [1979, 1996] and Zapata [2000].
• Test whether each pair of loci in each sample and overall is at genotypic
equilibrium
Fstat Test whether groups of samples differ for a large panel of statistics
Fstat Test whether categories of individuals differ in dispersal rates (four different
tests)
• Convert Fstat format to Genepop and vice-versa
Fstat Performs multiple regression or Partial Mantel tests
1.1.1 What’s new in version 2.9.4?

• made the help file a pdf document, printable and searchable.
• added the estimation of 2 coefficients of composite disequilibrium (R and ∆0 ).
• added the possibility of specifying a number of permutations for testing geno-
typic disequilibrium
• added the possibility of specifying a number of permutations for testing pair-
wise differentiation and HW in each sample.
• fixed a bug in the testing procedure of the partial mantel test.
todo generalized the bootstrap over loci as a way to obtain confidence intervals for
all the statistics.
todo added estimates of two genetic distances, Nei’s unbiased distance and Cavalli-
Sforza & Edwards Chord distance Nei [1987].
todo Hierarchical F-statistics
todo relatedness between pairs of individuals (Wang’s method (Wang [2002])).
todo Convert Fstat to Nexus format and vice-versa
todo Generates neighbor joining (NJ) trees of the samples, in the newick format
(compatible with e.g. Treeview (Page [1996])[http://taxonomy.zoology.
gla.ac.uk/rod/treeview.html]).
1.2 What the program FSTAT does NOT do

• Write your population genetics papers!
• Fstat is not suited for dominant data, obtained from e.g. RAPDs or AFLPs.
If you have such data, please have a look at Hickory(Holsinger et al. [2002])[http:
//darwin.eeb.uconn.edu/hickory/hickory.html]
• Fstat is not suited for sequence data. One of the suitable softwares for this
is Arlequin(Schneider et al. [2000])[http://lgb.unige.ch/arlequin/].
• Fstat will not assign individuals to populations. A very good software for
this is the Structure software(Pritchard et al. [2000])[http://pritch.bsd.
uchicago.edu/].
• check Genepop (Raymond and Rousset [1995b]) [ftp://ftp.cefe.cnrs-mop.
fr/genepop/] for testing for isolation by distance among individuals.
• The following web sites contain links to programs of interest for population
genetics analysis:
– Department of Biological sciences at Louisiana state university http:
//www.biology.lsu.edu/general/software.html
– Joe Felsenstein list at http://evolution.genetics.washington.edu/
phylip/software.html
Chapter 2
Running the program
The program is run by double clicking on the icon (a red butterfly) named Fs-
tat294.exe.
The first time you run the program, you will be prompted for 2 numbers, which
are the seeds of the Random Number Generator. Any 2 numbers will do. The
chosen generator is described in L’Ecuyer [1988]. It is a combination of 2 of the best
Multiple Linear Congruential Generator known, and has been thoroughly checked
(Goudet [1993]). In future run, the seeds will be loaded from the file Fstat.INI,
created when you exit the program and updated after each run. If for one reason
or another, you want to use a set of specific seeds, just edit the file Fstat.INI or
use the Utility, Reset seeds random number generator menu.
2.1 Creating an input file

For running Fstat, it is first necessary to create an input file named FILENAME.DAT
(where FILENAME is anything between 1 and 256 characters) containing the geno-
typic data, coded numerically, either with a 1, a 2 or a 3-digit number per allele.
The file must have the following format:
The first line contains 4 numbers: the number of samples, np(≤ 200), the number
of loci, nl(≤ 600), the highest number used to label an allele, nu(≤ 999), and a 1
if the code for alleles is a one digit number (1-9), a 2 if code for alleles is a 2
digit number (01-99) or a 3 if code for alleles is a 3 digit number (001-999). These
4 numbers need to be separated by any number of spaces. There is a further
constraint on the total number of individuals in the data set, which needs to be less
than 200’000.
The first line is immediately followed by nl lines, each containing the name of a
locus, in the order they will appear in the rest of the file.
On line nl + 2, a series of numbers as follow:
1 0102 0103 0101 0203 0 0303
The first number identifies the sample to which the individual belongs, the
second is the genotype of the individual at the first locus, coded with a 2 digits
number for each allele, the third is the genotype at the second locus, until locus
nl is entered (in the example above, nl = 6). Missing genotypes are encoded with
0. Note that 0001 or 0100 are not a valid format, that is, both alleles at a locus
have to be known, otherwise, the genotype is considered as missing. No empty
lines are needed between samples. The number of spaces between genotypes can be
anything. Numbering of samples need not be sequential, however, the number of
samples np needs to be the same as the largest sample identifier. Samples need not
to be ordered. nu needs to be equal to the largest code given to an allele , even if
8
CHAPTER 2. RUNNING THE PROGRAM 9
there are less than nu alleles (e.g. if only alleles 641 and 657 are present, nu = 657).
An example of 2 valid input files is given in Appendix A.
If you have a file with the Genepop format, Fstat now translate this format
into its own. See the File Conversion menu under Utilities
2.2 LIMITS IN FSTAT

• Maximum number of samples : 200
• Maximum number of loci : 600
• Maximum number of alleles : 999
• Maximum number of individuals : 200’000
It might be difficult memory-wise to run a data set with all these parameters to
the maximum, but I guess it would also mean quite some time reading gels!
If you need more capacity, please contact me : jerome.goudet@ie-zea.unil.ch.
Chapter 3
File menu
This is the only accessible menu when you start Fstat.
3.1 Open submenu

This is where to start! A window will open displaying all the files in the Fstat
directory with the default extension .DAT. If the data file is not in this folder, just
navigate using the ”look in” option of the window until you have located it. Once
you have it, double click the file, or click on open.
If many files with the same structure (that is, the same number of samples,
loci and individuals) need to be processed, they can all be selected at once1 , and
Fstat will process them with the default options sequentially. This is useful if you
want to test a scenario via simulations, using software such as Easypop (Balloux
[2001]http://www.unil.ch/izea/softwares/easypop.html).
Input files with an extension different from .DAT will not be opened.
3.2 Save Default Options subMenu

This menu allows you to start Fstat with your preferred options already selected.
You then just need to open a file, and click run!
3.3 Exit subMenu

Obvious, isn’t it?!
1 by clicking on the desired files while keeping the Ctrl key pressed
10
Chapter 4
Choose options menu
This menu is greyed as long as you have not chosen a data file (menu file, open).
Under this menu, you will select what you want Fstat to do. Some of the
options are selected by default (those with a tick mark). To select an unselected
option, or to deselect one, just click on it. Tick marks will then either appears or
disappears.
4.1 Label file for pops sub-menu

A text file containing names (labels) for the populations (samples) can be given
under this option. The default extension for label files is .LAB. Each line should
contain the name (label) of one sample, and samples should appear in the same
order as in the .DAT file. Labels can be any length but will be truncated to six
characters in the output files.
4.2 Loci to use sub-menu

It is sometimes useful to focus on only a subset of the collected data. For example,
if the study involves different classes of genetic markers such as microsatellites and
allozymes, it might be useful to analyse separately each class of markers. Another
potential application is if you identify an ill-behaved locus. This sub-menu will let
you pick the loci you want to focus on.
After a subset of the loci have been selected, when the run button is clicked,
you will be prompted for a file name under which to store the reduced data-set.
If you choose the default filename, it will be the original name, followed by -L
(for loci), followed by the numbers identifying the selected loci.
The chosen filename will be used for the input data as well as for all the output
files, with the appropriate extensions.
4.3 Samples to use sub-menu

It is sometimes useful to focus on only a subset of the collected data. For example,
if the study involves samples from different habitats, or regions, one might be in-
terested in focusing on one of these habitats. This sub-menu will let you pick the
samples you would like to focus on.
After a subset of the samples have been selected, when the run button is clicked,
you will be prompted for a file name under which to store the reduced data-set.
11
CHAPTER 4. CHOOSE OPTIONS MENU 12
If you choose the default filename, it will be the original name, followed by -P
(for populations), followed by the numbers identifying the selected samples.
The chosen filename will be used for the input data as well as for all the output
files, with the appropriate extensions.
The samples in the new file will appear with the order they had in the original
.dat file. That is, if you select (in this order) samples 5, 6 2 & 1, the new .dat file
will contains (in this order) samples 1, 2, 5 & 6. If a label file was associated with
the original .dat file, a new label file will be created, with labels corresponding to
the sampled populations (and in the right order1 ).
1 if you would like samples to be identified by a label, you have to first select the file containing
the labels, an only then to pick the samples. Proceeding the other way around (unless you have
already created a label file with the subset of samples you are interested in) will not associate
correctly samples with their respective labels.
Chapter 5
Per locus and sample

statistics
5.1 Allele frequencies

For each locus in each sample, will estimate the number of individual typed. The
allele frequencies in each sample and overall will then be estimated. For the overall
allele frequencies, both the weighted (by sample size) and non-weighted frequencies
are reported.
5.2 Genotypic frequencies

If you cheked this box, a file (extension .X2, see results files) containing the observed
and expected genotypic number for each genotype at each locus in each sample is
created. The expected numbers are unbiased estimates based on the hyper geomet-
ric distribution:
h i
i −1)
Ai Ai = n pi (2np
(2n−1) ,
h i
(2npj )
Ai Aj = 2n pi (2n−1)
where Ai Ai represents homozygote genotypes, Ai Aj represents heterozygotes, n

the sample size, pi and pj the allele frequencies of allele Ai and Aj respectively.
5.3 Number of alleles

Counts the number of different alleles observed at each locus in each sample, and
overall samples.
5.4 Allelic Richness

Estimates allelic richness per locus and sample (Rs ), and overall samples (Rt ).
Allelic richness is a measure of the number of alleles independent of sample size,
hence allowing to compare this quantity between different sample sizes.
The observed number of alleles in a sample is highly dependant on sample size.
To bypass this problem, El Mousadik and Petit [1996] suggested to adapt the rar-
efaction index of Hurlbert [1971] to population genetics (see also Petit et al. [1998]).
13
CHAPTER 5. PER LOCUS AND SAMPLE STATISTICS 14
The principle is to estimate the expected number of alleles in a sub-sample of 2n

genes, given that 2N genes have been sampled (N ≤ n). In Fstat, n is fixed as
the smallest number of individuals typed for a locus in a sample. Allelic Richness
is then calculated as:
" #
X 2N −Ni
2n
Rs = 1− 2N

2n
where Ni is the number of alleles of type i among the 2N genes. Note that each
term under the sum corresponds to the probability of sampling allele i at least once
in a sample of size 2n. If allele i is so common that we are certain to sample it
-when 2n > (2N − Ni )- the ratio is undefined but the probability of sampling the
allele is set to 1.
For Rt , the same sub-sample size n is kept, but N is now the overall samples
number of individuals genotyped at the locus under consideration. Rt is reported
in the last column.
Differences in allelic richness among groups of populations could be tested, see
tab-sheet Comp. Among groups of samples.
5.5 Gene diversities

Estimates gene diversity per locus and sample using the unbiased estimator
!
n k
X
HSk = 1− p2ki − HO
k
/2nk
nk − 1 i
(see [Nei, 1987, eq. 7.39 p. 164]).
5.6 Fis
k
k HO
Estimates FIS for each locus and sample as FIS = 1− k .
HS
Here, the generic name
FIS is given, rather than GIS or f (see below). This is because the statistic is
estimated for each sample, in which case the difference between GIS and f (due to
the different weightings of varying sample size, see below) vanishes.
Chapter 6
Global statistics
F-statistics are a set of tools devised by Wright [1921, 1969] to partition heterozy-
gote deficiency into a within and an among population component. They are widely
used by population biologists to assess levels of structuring in samples of natural
populations. FIS measures the heterozygote deficit within populations, FST among
populations (a measure of the Wahlund effect), and FIT the global deficit of het-
erozygotes.
Estimation of F-statistics has been debated in the literature for the past 30 years,
since the early work of Cockerham [1969, 1973] and Nei [1973, 1975]. Advantages
and problems of either estimator have been discussed at length (Slatkin and Barton
[1989], Chakraborty and Danker-Hopfe [1991], Cockerham and Weir [1993]). A good
review of these and other estimators can be found in Excoffier [2001].
Fstat calculates Nei and Weir & Cockerham estimators of gene diversities and
differentiation.
6.1 Nei’s estimators of F-statistics

The following statistics, defined in [Nei, 1987, eq. 7.38– 7.43 pp. 164–5] are esti-
mated:
• the observed heterozygosity
XX
HO = 1 − Pkii /np,
k i
where Pkii represents the proportion of homozygote i in sample k and np the

number of samples.
• the within population gene diversity1 :
" #
ñ X
2 H0
HS = 1− pi − ,
(ñ − 1) i
2ñ
where ñ = P np and p2i = p2ki /np

P
1/n k
k
k
• the overall gene diversity

X HS H0
HT = 1 − pi 2 + − ,
i
(ñnp) (2ñnp)
P
where pi = k pki /np.
1 sometimes misleadingly called expected heterozygosity
15
CHAPTER 6. GLOBAL STATISTICS 16
• the amount of gene diversity among samples DST = HT − HS

0 np
• DST = np−1 DST
• HT0 = HS + DST
0
DST 2
• FST = HT .
0
0 DST
• FST = HT
HO
• FIS = 1 − HS
Here, the pki are unweighted by sample size. These statistics are estimated for
each locus and an overall loci estimates is also given, as the unweighted average
of the per locus estimates. In this way, monomorphic loci are accounted for (with
estimated value of 0) in the overall estimates.
Note that the equations used here all rely on genotypic rather than allelic number
and are corrected for heterozygosity (see Nei and Chesser [1983]). A thorough
description of many estimators of gene diversity and differentiation is available in
the excellent review of Laurent Excoffier (Excoffier [2001])
6.2 Weir & Cockerham estimators of F-statistics

2
σa +σb2
Weir and Cockerham [1984] estimator of FIT , FST and FIS (F = 2 +σ 2 +σ 2 ,
σa
θ=
b w
σa2
σb2
2 +σ 2 +σ 2
σa
and f = respectively in Fstat output) are estimated for each
σb2 +σw2
b w
allele, locus and overall. For each locus and overall, Fstat will also output the
different variance components from which F-statistics are estimated, namely the
among sample variance component:
P P P
nk (pk −p̄)2 nk p̄(1−p̄)− nk (pk −p̄)2 −(1/4) nk HOk
np−1 − k kP
nk −np
k
2 k
σa = P 2
1
P n
k k
k nk −
P
np−1 n k
k
P
nk pki
where (p¯i = P
k
n
), the between individual within sample component
k
k
− p̄)2 − (1/4)
P P P P
nk p̄(1 − p̄) − nk (pk nk HOk nk HOk
σb2 = k kP k
− kP
k nk − np 4 k nk
and the within individual component

P
2 nk HOk
σw = kP .
2 k nk
For more details about these different quantities, refer to the original paper by
Weir and Cockerham [1984] as well as Weir [1996].
2 2
If all the samples have the same size, then σw is the same as HO , (σw +σb2 ) as
2 2 2 0 0
HS , (σw +σb +σa ) as HT , θ as FST and f as FIS . This is no longer true if sample
sizes are different. Weir & Cockerham (WC) weight allele frequencies according to
sample sizes (as would be done in an ANalysis Of VAriance), whereas Nei weights
all samples equally, whatever the sample size is. When sample sizes vary a lot, this
can lead to large differences between the two families of estimators.
2 This is not the same as Nei’s G
ST . Nei’s GST is an estimator of FST based on allele frequencies
only
Nei and WC also treat differently monomorphic loci: Under Nei’s family, when
0
a locus is completely monomorphic, FIS and FST (and FST ) are all 0, while WC
3
consider that the estimators cannot be defined .
A measure of Hamilton [1971] relatedness is also included, calculated using an
estimator strictly equivalent to Queller and Goodnight [1989] (see in particular
appendix B of this paper), namely: R = 2θ/(1 + F ). This measure is the average
relatedness of individuals within samples when compared to the whole.
Pamilo [1984, 1985] pointed out that when there is inbreeding, relatedness is
biased and proposed an ”inbreeding corrected relatedness”, labelled relatc in Fstat,
and estimated as: Rc = [R − 2F/(1 + F )]/[1 − 2F/(1 + F )] One should be wary
however, that this estimator is not bounded between 0 and 1. In particular when
working with species undergoing partial selfing, relatc seems to produce invalid
results. On the other hand, when the population is structured, relatc adequately
removes the increase in relatedness due to this structuring (see Chapuisat et al.
[1997] for an example)
2
Differences among groups for HO (σw ), HS (σb2 +σw
2
), f , θ, relat and relatc could
all be tested, see tab-sheet Comp. Among groups of samples.
6.2.1 Jackknife variance and Bootstrap confidence interval

Jackknifing over samples and loci and bootstrapping over loci are automatically
performed for the statistics F , θ, f and relat if the number of samples and the
number of loci are sufficient (more than 4). See Raymond and Rousset [1995a] for
a discussion of some problems encountered with bootstrapping (their argument is
buttressed on the data set diploid.dat given in the appendix) when there is not
a sufficient number of loci, and why randomisations are better suited.
6.3 RST
RST is an estimator of gene differentiation accounting for variance in allele size and
defined for genetic markers undergoing a stepwise mutation model (Slatkin [1995]).
It is not worth bothering about unless you are pretty sure that mutation follows
a stepwise mutation model quite strictly, and that mutation can not be neglected
compared to others forces such as migration (Balloux et al. [2000], Balloux and
Goudet [2002]). Fstat estimates RST for each locus following Rousset [1996].
This estimator does not depend on the number of samples. It also outputs the
different components of variance of allele size: Va for among samples, Vb for among
individuals within samples and Vw for within individuals. Vt = Va + Vb + Vw is thus
an unbiased estimate of the total overall variance in allele size.
Three estimators of overall loci RST are provided:
1. one according to Rousset [1996], where each locus is weighted by its amount
of allelic variance.
2. Because variances in allele size commonly vary by orders of magnitude among
loci, Fstat also gives the estimator of Goodman [1997]. Calling Mo and Vo
the mean and variance in allele size at a locus, Goodman
√ [1997] suggested
to use the centred normalised allele size (x − Mo )/ Vo instead of allele size
x in the estimation of RST . Individual loci RST will not be affected by this
centering-rescaling, but the overall loci estimator will, because each individual
3 it is sometimes found in the literature that F 0
ST or FST cannot be negative. While this is true
for the statistic GST because this statistic does not include a correction for heterozygosity (Nei
[1973]), it is wrong for subsequent versions accounting for heterozygosity. For details on this issue
and more about the comparison between Nei and WC estimators, Cockerham and Weir [1993]
locus variance component will be divided by its locus allelic size variance (since
s2 (aX + b) = a2 s2 (X)). Goodman’s RST will thus be expressed as:
Pl
Vai /Voi
U RST = Pi=1
l
i=1 Vti /Voi
where l represents the number

Pl of Loci. Vt is an unbiased estimator of Vo ,
so Vt ≈ Vo , and therefore i=1 Vti /Voi ≈ l (In fact, one could argue that Vt
should be used for the total variance instead of Vo ). This leaves us with :
l
X Vi a
U RST ≈ 1/l = RST U ,
i=1
Vti
the average of the individual Loci RST .

3. This last un-weighted estimate (RST U ) is also produced by Fstat for com-
pleteness.
For a review of the properties of these 3 estimators compared to FST , and some
advice on when to use them, see Balloux and Goudet [2002].
6.4 FST per pair

Calculates the multilocus Weir and Cockerham [1984] estimator of FST (θ) between
all pairs of samples. Two output files are produced. One with the extension .FST,
containing the table of pairwise θ. The other with the extension .MAT, containing
2 half tables. The first one gives, for each pair of samples, the sum of the three
variance components (σa2 , σb2 and σw2
). The second half table gives σa2 .
Since FST is closely related to a genetic distance (Reynolds et al. [1983]), the
first table can be used in Mantel tests (e.g., Manly [1985]). Rousset [1997] has
elucidated the appropriate transforms to be used on FST and geographic distances
if one wants to test for a linear association between these two quantities4 .
4 in the old days (v 1.2), Fstat produced a file containing the so-called ”Nm” values. These
values were simply a function of pairwise FST , namely 1/(4 FST ) - 1/4. In many cases however,
the assumptions necessary to transform FST into a number of migrants are not fulfilled (Whitlock
and McCauley [1999]). To avoid misuse, Fstat does not produce these values anymore
Chapter 7
Testing
7.1 Introduction
All the tests in Fstat are randomisation based. The principle behind randomisation
based tests is the following: Data sets fitting the null hypothesis to be tested are
generated by randomising the appropriate units (alleles, genotypes. . . see below).
A statistic is calculated on these randomised data sets and its value compared to
the statistic obtained from the observed data set. The proportion of statistics from
randomised data sets that give a value as large or larger than the observed provides
an unbiased estimation of the probability that the null hypothesis is true (P-value
of the test). These tests are carried out for each locus individually and then over
all loci. Bonferroni procedures should be applied when using individual population
tests or individual locus tests (Rice [1989]).
Units of randomisation are NOT the same for the different tests:
For HW within samples

Alleles are permuted among individuals, within samples. Then a statistic has to
be chosen to compare the randomised data sets to the observed. In Fstat, this
statistic is FIS (f ).
For HW overall
Alleles are permuted among samples. The statistic used to compare the randomised
data sets to the observed is FIT (F )
For population differentiation

Two types of tests can be carried out: If there is strong evidence that there is
HW within samples, then it is valid to permute alleles among samples to test for
population differentiation, since alleles can be considered as independent. On the
other hand, if HW is rejected within samples, alleles within an individual are not
independent anymore. In this case, permuting genotypes among samples is the only
valid permutation scheme1 .
In either case, contingency tables of alleles by samples are generated, and the
statistic that Fstat uses to classify them is the log-likelihood statistic G (Goudet
et al. [1996]):
1 Only complete multilocus genotypes are randomised, to make sure that the combined test
over loci is valid. If you’d like to use all possible single locus genotype to estimate the probability
of differentiation for individual loci, you’ll need to create separate file for each locus (using the
options, loci to use sub-menu)
19
CHAPTER 7. TESTING 20
np X
nu
X nik
G = −2 nik log ,
n k pi
k=1 i=1
where nik is the number of allele i in sample k, nk is the number of alleles (twice
the number of individuals) in sample k, and pi is the frequency of allele i in the
whole dataset for the overall test, and in the two samples considered for the pairwise
test (with np = 2 in this case). To test for population differentiation among all loci,
the following statistic is used to classify contingency tables:
np X
nl X nu
X nikl
G = −2 nikl log ,
nkl pil
l=1 k=1 i=1
where l indexes loci.

Two values are given for the P-values in the above options. These values corre-
spond to the proportion of randomised data that gave: (i) a value larger or equal
than the observed, (ii) a value strictly larger than the observed. These two values
are in general very close one to the other. But sometimes they are quite different,
and there are cases where they can cover the whole interval 0:1. These cases cor-
respond to loci with very little polymorphism, therefore bringing little information
about the population structure or the mating system. In any case, the P-value that
should be reported is the first one.
Test overall loci

The tests presented above are all designed for one locus. But how can one combine
the results of these tests to obtain a P-value overall loci? One option is to use
Fisher’s procedure to combine probabilities (Fisher [1954]) or more appropriately a
test based on the symmetry of the distribution of the P-values obtained from the
different loci (see Goudet [1999]). But both these tests give the same weights to the
different loci. One could wonder whether it is justified to give the same weight to a
bi-allelic, poorly polymorphic locus and to one highly polymorphic with 10 alleles.
An alternative is to construct a statistic accounting for all the loci, such as the sum
of the statistics obtained from individual loci, or a function of this sum, like the
mean. This is what Fstat does. The tests are carried out either on the sum or on
a weighted average of the statistic obtained from each locus (the weighting being
the same as the one used to estimate the overall loci statistic). In this way, loci
with a higher polymorphism receive more weights. So for the tests based on f , the
statistic used is the overall loci f , namely:
Pnl Pnu 2
σb
f = Pnl Pnu ,
(σb2 + σw
2)
where the first sum is over loci and the second over alleles. For the test based
on F , it is:
Pnl Pnu 2
σa + σb2
f = Pnl Pnu ,
(σa2 + σb2 + σw 2)
and for the tests of differentiation based on the likelihood ratio statistic, Fstat uses
as an overall test statistic the sum of individual loci G-values
np X
nl X nu
X nikl
G = −2 nikl log ,
nkl pil
l=1 k=1 i=1
A brief description of the performance of this last test is given in Petit et al.
[2001].
7.2 Global tests

7.2.1 Hardy Weinberg Within samples
Alleles are randomised among individuals, within samples. f is used to classify
contingency tables of alleles against alleles in each sample.
7.2.2 Hardy Weinberg Overall samples

Alleles are randomised over the whole dataset. F is used to classify contingency
tables of alleles against alleles in the whole dataset.
7.2.3 Population Differentiation NOT assuming Hardy Wein-

berg within
Rather than alleles, genotypes are randomised among samples. Contingency tables
of alleles against samples are obtained from the randomised data sets, and the
log-likelihood statistics G is used to classify them.
My advice is NOT to assume HW within samples for testing population differ-
entiation, unless you are certain that HW is realised within these samples.
7.2.4 Population Differentiation assuming Hardy Weinberg

within
Alleles are randomised over the whole data set. Contingency tables of alleles against
samples are obtained from the randomised data sets, and the log-likelihood statistics
G is used to classify them.
7.3 Tests per sample or pair of samples

7.3.1 HW tests per locus and sample
Alleles are randomised among individuals, within samples. FIS is used to classify
tables. This option will report 2 tables of P-values. The first table corresponds to
tests for deficit in heterozygotes, the second for excesses of heterozygotes.
The nominal level for multiple tests box allows you to define table wide levels
of significance at 5%, 1% or 0.1%, or to specify a given number of permutations.
7.3.2 Pairwise tests of differentiation

For each pair of samples, multi-loci genotypes are randomised between the two
samples. The overall loci G-statistic is used to classify tables2 . The nominal level
for multiple tests box allows you to define table wide levels of significance at 5%,
1% or 0.1%, or to specify a given number of permutations.
For pairwise tests of differentiation, The results are outputted to a file with the
extension -PP.PVL (for Per Pair P-VaLues). This file contains two tables. The first
2 This option does not produce individual loci pairwise P-values
gives the non-adjusted P-value for each pair. The second gives the pairwise signifi-
cance after standard Bonferroni corrections3 . ”***” corresponds to significance at
the 0.1% nominal level, ”**” significance at the 1% nominal level and ”*” signifi-
cance at the 5% nominal level. ”NS” stands for non-significant and ”NA” for not
available.
3 the reported significance levels are after strict (not sequential) Bonferroni corrections based
on the indicative adjusted P-value

Chapter 8
Composite disequilibrium
When two or more loci are present, it is of interest to verify whether alleles at the
different loci assort independently. The classical measure of gametic disequilibrium
is D = pAB − pA pB where pAB represents the frequency of gamete carrying allele A
at the first locus and B at the second, and pA and pB represents the frequency of
alleles A and allele B respectively (Lewontin and Kojima [1960]). With genotypic
data, estimation of gametic disequilibrium is impossible unless one is willing to
assume Hardy-Weinberg at each locus (Weir [1979]) because gametic type in the
double heterozygotes cannot be inferred. This is where the composite disequilibrium
is useful.
8.1 Estimating composite disequilibrium

∆AB is the classical estimator of composite disequilibrium described by Weir [1979,
1996]. Given 2 polymorphic loci, alleles A and B at the first and second locus
respectively, and calling Ā and B̄ the non A and non B alleles, we have the following
contingency table of counts for genotypes, with a total of n two locus genotypes:
AA AĀ ĀĀ
BB n1 n2 n3
B B̄ n4 n5 n6
B̄ B̄ n7 n8 n9
The double heterozygotes (n5 ) can be produced by 2 types of pairs of gametes:

(AB)(ĀB̄) and (AB̄)(ĀB). It is therefore not possible to estimate the proportion
of gametes AB, but it is straightforward to obtain the count of (AB) gametes plus
the count of ”non gametic” (i.e one allele residing in each parental gamete) (A/B).
This sum is simply nAB = 2n1 + n2 + n4 + n5 /2. If the 2 loci assort independently,
the expected proportion of gametic and non-gametic (AB), (A/B) is 2pA pB . This
leads to the following estimator of composite linkage disequilibrium:
nAB
∆AB = − 2pA pB
n
An unbiased estimator of ∆AB is ∆d AB = n/(n − 1)∆AB (Weir [1996]).

∆AB , like its gametic equivalent DAB suffers from a major drawback: it is
bounded between [−2 × min(pA pB , pĀ pB̄ ), 2 × min(pA pB̄ , pĀ pB )]. In order to make
the range of the composite genotypic disequilibria less dependent on allelic frequen-
cies, the following statistic, similar to the D0 measure of gametic disequilibrium
(e.g. Hedrick [2000]), can be defined:
23
CHAPTER 8. COMPOSITE DISEQUILIBRIUM 24
∆AB
∆0AB = , ∆AB < 0
2 × min(pA pB , pĀ pB̄ )
∆AB
∆0AB = , ∆AB ≥ 0
2 × min(pA pB̄ , pĀ pB )
Weir [1979, 1996] suggested another standardization, leading to a coefficient akin

to a correlation coefficient, hence its name RAB :
∆AB
RAB = p
(pA pĀ + DA )(pB pB̄ + DB )
where DA = nAA 2 nBB 2
n − pA and DB = n − pB represent excess homozygosity at
locus A and B respectively. (pA pĀ + DA ) is the expression for the variance of the
frequency of allele A given its frequency pA and departure from HW DA . Hence,
the square root of he product of this quantity for alleles A and B is the product of
the standard deviation, and since ∆AB is the covariance of the frequency of allele
A and B, RAB is indeed akin to a correlation coefficient.
When the loci are multi allelic, the number of disequilibrium values per pair
of loci is n ∗ m, with n alleles at the fist locus and m at the second. It seems
worthwhile to combine these into a single measure for the locus. A natural multi-
allelic estimator consists in taking the weighted average of the |∆0AB |, where the
weight is the expected frequency of the gamete, pA pB (really, it should be 2pA pB ,
dividing the total by 2):
n X
X m
∆0 = pi pj |∆0ij |
i=1 j=1
The statistical properties of multiallelic ∆0 have not been investigated, but those
of the equivalent estimator of gametic disequilibrium D0 have been thoroughly in-
vestigated by Zapata [2000], Zapata et al. [2001]. In particular, Zapata et al. [2001]
showed that under many conditions, the distribution of multi-allelic D0 do not de-
viate from normality.
An overall samples estimator of ∆0 is also provided1 as:
Pnp 0
k=1 nk ∆k
∆0 = P np
k=1 nk
I am unaware of multi-allelic estimators of R2 . Following the logic used to obtain

one for ∆0 , I suggest the following:
n X
X m
R2 = 2
pi pj Rij
i=1 j=1
An overall samples estimator of R2 is also provided:

Pnp 2
k=1 nk Rk
R2 = P np
k=1 nk
If check-box Delta’ is checked, values of ∆0 estimated for each pair of locus in

each sample are saved in a file with the extension -dp.cd.
Similarly if check-box R^2 is checked, values of R2 estimated for each pair of
locus in each sample are saved in a file with the extension -r2.cd.
1 perhaps a better weight would be n H k H k , where H k
k Sl1 Sl2 Sl1 represents gene diversity at the
first locus in the kth sample
If check-box detail results is checked, a file with the extension -details.cd

is created. It contains, for each sample and and pair of alleles for each pair of locus
the following information:
• N , the number of 2 locus genotypes in the sample for the pair of locus con-
sidered
• pA , the frequency of the allele considered at the first locus
• pB , the frequency of the allele considered at the second locus

• cA = pA (1 − pA ) + PAA − p2A , where PAA is the frequency of homozygote AA.
• cB = pB (1 − pB ) + PBB − p2B
• nAB = 2n1 + n2 + n4 + n5 /2
n
• ∆d
AB = n−1 (nAB /n − pA pB )
• ∆0AB
• RAB
Last, it might be useful to have a list of all the contingency tables of genotypes
by genotypes. If you check the box Save contingency Tables, a file with the
extension -tables.ld will be created. It can be quite large, as it will contain for
each pair of loci in each sample the cross-table of 2 locus genotypes.
8.2 Testing for genotypic disequilibrium

This option allows testing the significance of association between genotypes at pairs
of loci in each sample2 . The statistic used to test the tables is the log-likelihood
ratio G-statistic3 . Clearly, only individuals typed at both loci enter the table.
The P-value of the test is obtained as follows. Genotypes at the 2 loci are
associated at random a number of times and the statistic is recalculated on the
randomised data set. The P-value is estimated as the proportion of statistics from
randomised data sets that are larger or equal to the observed.
An overall sample statistic is obtained by summing (over samples) the G-statistics
obtained in each sample. The overall test is obtained by comparing this overall
statistic with that obtained from randomised tables (randomisation occurring of
course only within samples).
If the option Tests between all pairs of loci in each sample is chosen,
then the P-value for each pair of loci in each sample as well as overall is produced.
The number of randomisations is fixed by the nominal level for multiple test radio-
group box. It will be 20 × np × nl × (nl − 1)/2, 100 × np × nl × (nl − 1)/2 or 1000 ×
np × nl × (nl − 1)/2 randomisations for a 5%, 1% or 0.1% nominal level, respectively
(by nominal, I mean at the desired level after Bonferroni corrections). With 10
samples and 10 loci, these numbers are 9’000, 45’000 and 450’000 respectively.
For polymorphic markers such as microsatellites, these permutations can take an
exceedingly long time.
2 This test correspond to hypothethis P in Zaykin et al. [1995]: Two-locus genotype frequency
4
is product of two one-locus genotypes frequencies. It is therefore valid when the two loci are not
in HWE
3
Pnmore Pmaccurately, the only part of this statistic that changes when randomising tables:
i≤j
x
k≤l ijkl
log (xijkl ), where xijkl represents the number of individuals in the sample with
genotype ij at the first locus and genotype kl at the second locus
In most case you are only interested in overall samples level of significance. You
should then choose the option Tests between all pairs of loci. The number
of randomisations for this option is fixed by the nominal level for multiple test radio-
group box. It will be 20×nl×(nl−1)/2, 100×nl×(nl−1)/2 or 1000×nl×(nl−1)/2
randomisations for a 5%, 1% or 0.1% nominal level, respectively. With 10 loci,
these numbers are 900, 4’500 and 45’000 respectively. It can still take some time
for loci with a large number of alleles, as for each randomisation, the test needs to
reconstruct the cross-table. With 20 alleles at the 2 loci for instance, the dimension
of the table to scan is 210 × 210!.
Because the number of randomisations can turn out to be very large, I added
the possibility to fix this number, using the fixed number option combined with
the number of permutations box (1000 by default).
Chapter 9
Run button
This button is greyed as long as you have not chosen a data file
Once you have selected all the options you want, just click here to run the
analysis. The results are written (appended if the file already exists) to the .OUT
result files.
If you have selected a subset of samples and/or loci, when you click on this
menu, Fstat will prompt you for an alternative filename for the subset of data.
The default is the original filename, followed by ”-L” and numbers corresponding
to the selected loci, followed by ”-P” and numbers corresponding to the selected
samples. It is a good idea to change this default filename, that could turn out to be
very long, and not very meaningful after a few weeks without using the data set!.
27
Chapter 10
Comp. Among groups of

samples
Under this tab-sheet, tests for difference among groups of populations for a
number of statistics (allelic richness, Observed heterozygosity, Gene diversity, FIS ,
FST , relatedness and corrected relatedness) can be carried out. The number of
groups can be anything between 2 and 12. Each group needs to be made of at least
2 samples. If only two groups are compared, the tests could be one or two sided,
otherwise, they are two sided.
28
CHAPTER 10. COMP. AMONG GROUPS OF SAMPLES 29
10.1 Principle of the tests

Fstat first calculates for each group the average (over samples and loci) of the
chosen statistic x (where x could be allelic richness Rs , Observed heterozygosity
HO , Gene diversity HS , FIS , FST , relatedness or inbreeding corrected relatedness).
If only 2 groups are chosen and a one sided probability is requested, then the
difference in x between group 1 and group 2 (x1 − x2 ) is taken. Otherwise, the
following quantity is calculated:
ng−1
X ng
X
OSx = (xi − xj )2 ,
i=1 j=i+1
where ng represents the number of groups. To assess the significance of the

statistic OSx , a permutation scheme is used. Whole samples are allocated at random
to the different groups (keeping the number of samples in each group constant), and
Sx is calculated from the randomised data set. It is important to remember that for
these tests, the unit of observation is the sample, not the individual. The P-value
of the test is then taken as the proportion of randomised data sets giving a larger
Sx than the observed OSx .
10.1.1 Number of groups

This is the first thing you need to define. This number is constrained to be between
2 and 12.
10.1.2 Define groups

On clicking this button, a window containing 2 lists appears. The list on the left
contains all the samples. If you choose to give a label to your different samples
(menu option, label file for pops) then their name appears. Select the samples (at
least 2, otherwise the test will be meaningless) belonging to the first group, click
on the ”>” sign, and the samples selected will move from the list on the left to
the list on the right. Click OK. The same window reappears to define the second,
and subsequent groups. Be careful not to select the same sample for two different
groups! If you made a mistake, click the ”Cancel” button.
10.1.3 Type of tests for two groups

If only 2 groups were chosen, then a radio group box appears next to ”number
of groups” after you’ve define the 2 groups. You could choose here whether you
want to test the null hypothesis of no differences against the alternative hypothesis
x1 > x2 , or against a two sided alternative.
10.1.4 Statistics to compare among groups

Allelic richness RS is calculated as described in the per locus and sample statistics
section. RSk for group k is taken as the (non weighted) average over loci and samples
of group k.
Observed heterozygosity, gene diversity, FIS , FST , relatedness and corrected re-
latedness are calculated as described in the Global statistics section. These statistics
are all weighted by sample size.
CHAPTER 10. COMP. AMONG GROUPS OF SAMPLES 30
10.1.5 Number of permutations

The minimum is 200, the maximum 15’000. Try 1’000 first (it can take very long,
particularly if allelic richness is estimated for a large sub-sample). If the P-value
you’ve obtained is close to the nominal level, increase the number of permutations.
10.1.6 Run
Run the tests. Results will be stored in a file with the name filename_test.out.
Each time you run a new test, results are appended to this file.
Hint: the units of randomisation for these tests are the samples. To increase
Degrees of freedom (and therefore the tests’ power), it is better to compare many
samples with perhaps fewer individuals rather than the opposite. I guess if you are
using Fstat now, it might be already too late to change the sampling design.
Chapter 11
Biased dispersal
11.1 Introduction
This option tests for biases in dispersal among two apriori defined groups of individ-
uals using information from codominant genetic markers. See Goudet et al. [2002]
for details of the principle and methods, and a power analysis of the different tests.
This option can also be used to test for other things than dispersal such as different
levels of inbreeding (using the observed heterozygosity HO , see below), colour, size,
parasitic state etc. . . .
11.2 Design of the tests

The tests assume an animal species with non-overlapping generations where dis-
persal occurs at the juvenile stage, before reproduction. They further assume that
individuals are sampled post dispersal. Under these conditions, several statistical
descriptors of an individual’s genotype can be used to indicate biases in dispersal.
In what follows, I use the example of sex-biased dispersal, but any other grouping
would work as well.
11.2.1 Assignment index AI

This statistic was first introduced in Favre et al. [1997]. For each individual j in
locality k, the program calculates the probability Pkij that its genotype at locus
i should appear in the kth sample as the squared frequency of the allele if the
individual is homozygous, or twice the product of the frequencies of its two alle-
les if it is heterozygous. If the loci segregate independently, then the probability
of occurrence of a multi-locus genotype is the product of the probabilities of the
individual loci. Because populations can contain very different levels of gene di-
versity, the multi-locus probabilities of individuals in different populations are not
directly comparable. To remove this problem, the average probability of the sample
is subtracted from the individual multi-locus probability, after log-transformation
to avoid rounding errors with very small numbers. This gives the following formula
for AIc , the corrected assignment index of individual j in sample k:
" l # n
" l #
Y X Y
AIckj = log Pkij − 1/n log Pkij
i=1 j=1 i=1
for l loci and n individuals in the kth sample. The distribution of AIc will
therefore be centered around 0. A positive value indicates a genotype more likely
31
CHAPTER 11. BIASED DISPERSAL 32
than average to occur in its sample (likely a resident individual), while a negative
value indicates a genotype less likely than average (potentially a disperser).
mean of AIc (mAIc ) Because immigrants tend to have lower AIc values than
residents, under sex biased dispersal, the average index for the sex that disperses
most is expected to be lower than that for the more philopatric sex. A t-statistic
mAIcp −mAIcd
is used for the test: t = q ,where np and nd are the number of
s2 p /np +s
2
d
/nd
AIc AIc
individuals in the more philopatric and the more dispersing group respectively.
variance of AIc (vAIc ) Because members of the dispersing sex will include both
residents (with common genotypes) and immigrants (with rare genotypes), vAIc for
the sex dispersing most should be largest.
11.2.2 FST
FST is a statistic expressing the proportion of the total genetic variance that resides
among populations (Hartl and Clarck [1997]). Allelic frequencies for individuals of
the sex dispersing most should be more homogeneous than those for individuals of
the more philopatric sex. We therefore expect FST for the more philopatric sex to
be higher than that of the more dispersing sex. Among the available estimators of
FST , we choose Weir and Cockerham [1984], because it is the most commonly used
and it is also unbiased.
11.2.3 FIS
FIS is a statistic describing how well the genotype frequencies within populations
FIT with Hardy Weinberg expectation (Hartl and Clarck [1997]). If only males
disperse, the males sampled from a single patch will be a mixture of two populations,
residents and immigrants; due to the Wahlund effect, the sample should show a
heterozygote deficit and a positive FIS . In general, members of the dispersing sex
should therefore display a higher FIS than the more philopatric sex. Among the
several estimators of FIS , we also choose Weir and Cockerham [1984].
Relatedness
relatedness is related to FST as relat=2 FST /(1+ FIT ). A test based on relatedness
has essentially the same properties as one based on FST , and is provided here for
convenience.
11.2.4 HO
The observed heterozygosity, HO , is NOT expected to change with the dispersal
status (nor is the total gene diversity HT , providing it is calculated with a weighting
proportional to the size of each group). But HO should differ among inbred and
outbred individuals. It might be of interest to test whether an individual’s status
is linked to inbreeding. For instance, one could test whether parasitized individuals
tend to be more inbred than healthy ones. If you have such categories, then a test
based on HO is of interest. See Trouvé et al. [2003] For an application of this test.
11.2.5 HS
The within group gene diversity HS should be largest for the group dispersing most.
11.2.6 Testing
To test whether these statistics differ significantly between the two sexes, a ran-
domisation approach is used. Under the null hypothesis that males and females
disperse equally, the four statistics do not depend on the variable ’sex’. Letting
Xd and Xp be the statistic of interest for the dispersing and the philopatric sex
respectively, the program proceeds as follows for one sided tests.
1. It first calculates the statistic for each sex over all populations and either take
the difference for FIS , HS and HO (for HO the group supposed to be most
outbred is assimilated to the group dispersing most); the t-statistic defined
above for mAIc ; the difference for FST and relatedness; or the ratio for vAIc .
2. It randomly assigns a sex to each of the multi-locus genotypes (keeping the
genotypes in their original sample, and the sex ratio in each sample constant).
3. It recalculates the appropriate difference or ratio for the randomised data set.
4. steps 2/ to 3/ are repeated numbperm times.
The probability that dispersal is unbiased by sex is then estimated as the propor-
tion of times where the relevant statistic is larger or equal to the observed one. The
two-tailed test are constructed under the same principle using either the absolute
value of the differences, or the ratio of the largest to smallest variance.
11.3 Running the program

11.3.1 Input file format
It is a slight modification of the Genepop format (Raymond and Rousset [1995b])1 .
Here is a brief exemple:
5 8 64
loc-3ks, loc-17, loc-3T7, loc-57, loc-72, loc-23, loc-9, loc-53
Pop
F,0, 4041 2627 3234 2828 2830 2122 2223 2022
M,0, 3940 2929 3232 3333 2430 2022 2329 0
M,0, 4040 2727 3232 2833 2731 0 2222 2020
F,0, 4059 2429 3434 3233 2430 1920 2325 2323
F,0, 4059 2429 3232 3233 2430 1920 2329 2023
F,0, 4040 2429 3232 3233 3030 0 0 2222
F,0, 4041 2829 3234 2931 2428 1919 2223 2023
M,0, 4059 2930 3434 3233 2425 1920 2531 2020
M,0, 4040 2429 3234 3233 0 0 2323 2023
Pop
F,0, 4041 2629 3334 3031 2428 2021 2326 2021
F,0, 4040 2735 3434 2833 2528 2023 2326 2020
M,0, 4040 2929 3333 2931 2424 2022 2223 2020
M,0, 4040 2629 3334 3030 2425 2020 2326 2121
F,0, 3940 2929 3234 2932 2425 1919 2326 2023
M,0, 4040 2729 3333 2931 2430 0 2323 2023
M,0, 4040 2626 3434 2931 2830 1921 2326 2021
F,0, 4040 2629 3334 3031 2425 2020 2223 2121
F,0, 4041 2727 3234 3232 2428 2021 2326 2123
M,0, 4040 2729 3233 3031 2430 2022 2223 1920
Pop
F,0, 4040 2729 3234 2933 2425 1919 2223 2023
M,0, 4040 2727 3434 2929 2425 1919 2223 1821
M,0, 4041 2729 3434 0 2528 1922 2123 2323
M,0, 4040 2727 3234 2933 2527 1919 2326 1821
F,0, 4040 2429 3333 2929 2324 2223 2323 1818
M,0, 4040 2929 3232 2933 2427 1919 0 2121
M,0, 4040 2929 3434 2933 2525 1919 2326 2121
M,0, 4041 2727 3434 2932 2528 1919 2326 2121
M,0, 4041 2929 3434 3333 2528 1919 2226 2121
The format has to be the following:
• The file name has the extension .GEN. No other extension will be recognised
by the program.
• The first line contains 3 numbers:
1 See menu utilities, file conversion for creating Genepop format using Fstat
– the number of samples,

– the number of loci, and
– the highest index used for an allele.
Alleles are to be encoded as 2 number digits (from 01 to 99).
• The second line contains the name of the loci, separated by comas.
• The third line contains the keyword ’Pop’ indicating the start of a new sample.
• The following lines contains the multilocus genotypes for each individual, pre-
ceded by a capital letter defining the group to which it belongs, a comma,
another descriptor of the individual (could be any string of characters exclud-
ing a comma) and a second comma. The capital letters defining the group
could be anything but a ’P’, the program will generate an error message if
you use a ’P’.
• Each new sample is separated from the previous by the keyword ’Pop’. See
the example of input data set.
11.3.2 Running the program

Click on the biased dispersal menu
Here is a screen shot of the program windows. To run the program, do the
following:
1. choose the file containing the data under the File, open menu.
2. type in each box (Letter for the most philopatric group and Letter
for the other group) the letter used in the data file to define members of
the two groups
3. choose whether you want a one or a two-sided test
4. choose which test(s) statistic you want to use
5. enter the number of randomizations for the tests (between 100 and 10’000)
6. click the GO button.
11.3.3 Results
The results are stored in a file with the extension .res and the same name as the
input file. An example of such a file is given in Appendix B. The program also
produces 4 other files:
• One is with the extension .per and contains the results of all the permutations.
The first line of the file contains the observed values for the differences between
groups of the different statistics (in the order Mean ass, Var ass, FST and FIS ,
relatedness, HO and HS ), preceded by the number 0. The remaining lines
contain the same information for each of the (numbperm-1) randomisations.
With this file, you can generate the distribution of the statistics under the
null hypothesis, as is done in the article describing the tests (Goudet et al.
[2002]).
• Three are files with the extension .dat and with the data reformatted so that
they can be analysed with Fstat . One of these files contains the whole
data set, the two others contain the data for each group, with the letter
corresponding to the group identifier appended to the file name.
Chapter 12
Mantelize it!
Under this menu, one can carry out multiple regression or partial Mantel test. This
latter option is available only providing the data originated from distance matrices.
While Raufaste and Rousset [2001] have criticized partial mantel, these tests are
still useful (rewrite)
12.1 Matrices to Columns

In order to perform a multiple regression or a (partial) mantel test with Fstat
(see menu multiple regression and partial mantel), the data must be in a columnar
mode. But often the data from distance matrices to perform mantel tests are in
matrix form, this is for instance the case of pairwise FST values produced by Fstat.
The menu ’Matrices to Columns’ allows you to combine the data from different files
in matrix form (whole matrices, with 0 on the diagonal) into a single file with the
data in columnar form (the half matrix below the diagonal is read, implying that
datum [2,1] is first read, followed by data [3,1] & [3,2] etc. until datum [n,n-1] where
n is the dimension of the matrices). When prompted for ”File names for data in
matrix form”, select all the files containing data you want to see appearing in the
multiple regression or partial mantel analysis. Clearly, these files need to have the
same dimension (same number of lines and columns). Multiple selections can be
achieved by holding the Ctrl key while clicking on the different files. The program
then prompts the user for the name of the columnar mode file. No specific extension
is required for this file. An example of such a file is given in the distribution. See
Example of input file for multiple regression and partial mantel
12.2 Multiple regression and partial mantel

This is an adaptation of Manly [1997] code for multiple regression, described in the
chapter ”regression analysis” of the book. The modification I made only concerns
analyses with data taken from symmetrical matrices (pairwise FST values for in-
stance). If you have such data, the program detects it (by counting the number of
observations: if the data are from a distance matrix, then they should be n(n − 1)/2
observations in the file). This will then modify the type of randomisations to be
carried out when testing significance of the multiple regression.
Once you have chosen a file with data in the appropriate format (Example of
input file), you need to first pick the dependant variable, usually a genetic distance.
Once it is chosen, you can pick up to 10 explanatory variables. If the number
of observations is compatible with the data being taken from a distance matrix,
a check-box appears allowing you to specify the appropriate permutation scheme.
36
CHAPTER 12. MANTELIZE IT! 37
You also have the possibility to save residuals in a file. The default number of
permutations for the test is set at 2000. Once everything is selected, click on run.
The results appear in the window at the lower right hand of the panel, while the
two upper right panels show scatterplots of the estimates and residuals, and fitted
and observed values. By left-clicking the graph, you will see the residuals (right
panel) and the estimates (left panel) as a function of the estimates (right panel) or
and the observed values (left panel).
The result file is structured as follows:
A brief description of the input file is given (name and comments). Then follows
the number of randomisations used in the test.
The total sum of square of the dependant variable is then given, followed by the
result of the regression: the (partial) correlation of each explanatory variable with
the dependant variables, the coefficient associated with each explanatory variable
and the sum of squares explained by the explanatory variables.
The overall percentage of the variance explained by the model (R-squared is
then given).
Last, the P-value for the coefficient associated with each variable is given, to-
gether with the P-value associated with the proportion of variance explained by each
explanatory variable. These P-value are given as percentage, that is, a 5 would mean
5% of the randomisation gave values as large or larger than the observed.
Each time a new test is run, it is appended at the end of the result file.
Chapter 13
Utilities menus
13.1 Reset seeds Random number generator

A dialog box appears, allowing you to reset the seeds of the random number gener-
ator. This is equivalent to editing the contents of the file FSTAT.INI
13.2 File conversion

This menu allows you to convert Fstat format to GENEPOP and vice-versa.
GENEPOP data files need to have the extension .GEN in order to be recognised
by Fstat.
The conversion from GENEPOP to Fstat will not work if the label of an in-
dividual (the text before the comma on each line of the GENEPOP format) starts
with the characters ”POP”, ”pop” or any case combination of these three letters.
If the label starts with the letter ”P”, make sure that it is longer than 3 char-
acters (spaces are valid characters, so the label ”PXX”, where X stands for a space
or any other character [0..9; a..z; A..Z], is valid).
38
Chapter 14
Files of results
Results will be stored in a file named FILENAME.OUT, located in the same folder as
the input file.
This file first contains a line saying when the analysis has been run. On the
following lines, one find the identifier of the sample, then, for each locus, the size
of each sample, followed by the frequency per sample and overall frequency of each
allele present at the locus. The rest of the output file should be self explanatory.
If you requested that pairwise FST be estimated under the menu choose options,
F-statistics, FST per pair 2 files are created:
• A file FILENAME.FST. This file contains the estimated pairwise FST
• A File FILENAME.MAT. It contains two half matrices. The first one corresponds
to the pairwise sum of variance components (σa2 , σb2 and σw
2
) representing the
denominator of Cockerham estimator of FST . The second corresponds to the
pairwise variance component σa2 , the numerator of Cockerham estimator of
FST .
If you requested that genotypic frequencies be calculated under the menu choose
options, gene diversities, genotypic frequencies, a File FILENAME.X2 will be cre-
ated. It contains for each genotype at each locus the observed and expected geno-
typic count, where the expected genotypic count is calculated using unbiased esti-
mation:
h i
i −1)
Ai Ai = n pi (2np
(2n−1) ,
h i
(2npj )
Ai Aj = 2n pi (2n−1)
where Ai Ai represents homozygote genotypes, Ai Aj represents heterozygotes, n

the sample size, pi and pj the allele frequencies of allele Ai and Aj respectively.
If you have selected the sub-option, Testing, Pairwise Pop. diff. NOT assuming
HW within submenu, a file FILENAME-PP.PVL containing two tables is generated.
The first gives the non-adjusted P-value for each pair. The second gives the pairwise
significance after standard Bonferroni corrections. ”***” corresponds to significance
at the 0.1% nominal level, ”**” significance at the 1% nominal level and ”*” signif-
icance at the 5% nominal level. ”NS” stands for non-significant and ”NA” for not
available.
If you have requested to estimate the composite disequilibrium ∆0 , a file with
extension -dp.cd will be created. Similarly, if you have requested the composite
disequilibrium R2 , a file with extension -r2.cd will be generated. If you requested a
39
CHAPTER 14. FILES OF RESULTS 40
detailed output, a file -details.cd will be generated. Last, if you requested Fstat
to save contingency tables, a file with extension -tables.ld will be generated.
All the separators in the output files are tabs, which allows direct importation
of the results into commercial packages such as the Microsoft spreadsheet Excel,
therefore facilitating printing and graphical representation of the results.
Chapter 15
Special use of Fstat
15.1 Haploid data

While Fstat has been written for diploid data sets, it could also be used for haploid
species. To do so, it is necessary to encode haploid genotypes as diploid genotypes,
by writing twice the identifier of the allele. A brief example follows:
3 4 4 1
loc-1
loc-2
loc-3
loc-4
1 11 22 11 11
1 22 22 11 22
1 11 33 11 22
1 22 22 33 11
2 11 11 11 11
2 11 22 11 22
2 11 11 11 11
3 22 11 11 11
3 33 11 11 11
3 22 11 33 11
Note that all genotypes are homozygous. When running this file, FIS (f ) and
FIT (F ) will be meaningless. FST (θ) will however be an appropriate measure of the
Wahlund effect. The file HAPLO.DAT, given in Weir [1996] ’Genetic Data Analysis’,
is distributed with the example files.
CAUTION : one should not mix haploid and diploid data in the same data file.
If you have autosomal markers and mitochondrial or Y chromosome markers, do
not put them in the same file but rather create two files, one for the autosomal loci,
the other for the haploid loci. The same goes for haplo-diploid species. This is to
insure that your overall loci statistics have some meaning!
15.2 Only one sample

In this case, Fstat will output the statistics it can estimate, namely allele frequen-
cies, gene diversities, number of alleles sampled, FIS if requested, test of FIS if
requested, as well as genotypic disequilibrium statistics and tests.
15.3 Pooling samples

One may want to calculate F-statistics for different levels of grouping of samples,
and follow the changes of, e.g. FIS (f ) with different levels of pooling. An example
of such a procedure is given in Goudet et al. [1994], Goudet [1993]. To allow for this,
one needs to modify the input file by renumbering samples according to pooling,
and modifying the number of samples np on the first line.
41
CHAPTER 15. SPECIAL USE OF FSTAT 42
15.4 Missing data

The situation may arise where some loci could not be screened for all the samples.
Fstat handles this situation properly. But I can only urge users to avoid data sets
with many missing genotypes. In particular when some samples have no individuals
typed at one or more loci, you might run into all sorts of problems. You’ve certainly
heard from statistical courses that balance sampling is the best. Well, It really is!
Appendix A
Examples of input files
One digit encoding for data file Two A.2 Two digits encoding
digits encoding for data file Label file
for input data file
The same file with alleles encoded with
A.1 One digit encoding a 2 digit number:
for input data file 6 5 4 2
loc-1
loc-2
A file encoded with 1 digit number (This loc-3
loc-4
is the example file DIPLOID.DAT, given loc-5
1 0404 0403 0403 0303 0404
in Weir’s (1990) ’Genetic Data Analysis’. 1 0404 0404 0403 0303 0404
1 0404 0404 0403 0403 0404
1 0404 0404 0 0303 0404
6 5 4 1
1 0404 0404 0204 0304 0404
loc-1
1 0404 0404 0 0403 0404
loc-2
1 0404 0404 0403 0403 0404
loc-3
1 0404 0404 0 0403 0404
loc-4
2 0404 0404 0303 0302 0404
loc-5
2 0404 0303 0404 0403 0404
1 44 43 43 33 44
2 0404 0403 0404 0403 0404
1 44 44 43 33 44
2 0404 0404 0303 0303 0404
1 44 44 43 43 44
2 0404 0403 0404 0404 0404
1 44 44 0 33 44
2 0404 0404 0404 0202 0404
1 44 44 24 34 44
2 0404 0404 0403 0403 0404
1 44 44 0 43 44
2 0404 0404 0404 0404 0404
1 44 44 43 43 44
3 0404 0404 0404 0403 0404
1 44 44 0 43 44
3 0404 0404 0404 0404 0404
2 44 44 33 32 44
3 0404 0404 0403 0201 0404
2 44 33 44 43 44
3 0404 0404 0303 0403 0404
2 44 43 44 43 44
3 0404 0404 0403 0201 0404
2 44 44 33 33 44
4 0404 0404 0403 0404 0404
2 44 43 44 44 44
4 0404 0404 0403 0403 0404
2 44 44 44 22 44
4 0404 0404 0403 0403 0404
2 44 44 43 43 44
4 0404 0404 0403 0404 0404
2 44 44 44 44 44
4 0404 0404 0403 0404 0404
3 44 44 44 43 44
4 0404 0404 0404 0303 0404
3 44 44 44 44 44
4 0404 0404 0404 0404 0404
3 44 44 43 21 44
5 0404 0404 0404 0201 0404
3 44 44 33 43 44
5 0404 0404 0404 0303 0404
3 44 44 43 21 44
5 0404 0404 0403 0403 0404
4 44 44 43 44 44
5 0404 0404 0403 0403 0404
4 44 44 43 43 44
5 0404 0404 0404 0404 0404
4 44 44 43 43 44
5 0404 0404 0404 0403 0404
4 44 44 43 44 44
5 0404 0404 0403 0403 0404
4 44 44 43 44 44
5 0404 0404 0404 0 0404
4 44 44 44 33 44
5 0404 0403 0404 0403 0404
4 44 44 44 44 44
6 0404 0404 0404 0403 0404
5 44 44 44 21 44
6 0404 0404 0403 0303 0404
5 44 44 44 33 44
6 0404 0404 0404 0302 0404
5 44 44 43 43 44
6 0404 0404 0403 0401 0404
5 44 44 43 43 44
6 0404 0404 0404 0404 0404
5 44 44 44 44 44
6 0404 0404 0404 0402 0404
5 44 44 44 43 44
6 0404 0404 0404 0403 0404
5 44 44 43 43 44
5 44 44 44 0 44
5 44 43 44 43 44
6 44 44 44 43 44
6 44 44 43 33 44
6 44 44 44 32 44
6 44 44 43 41 44
6 44 44 44 44 44
6 44 44 44 42 44
6 44 44 44 43 44
43
APPENDIX A. EXAMPLES OF INPUT FILES 44
A.3 Example of label file A.5 Example of input file

A file of labels for the samples in diploid.dat. for multiple regres-
Each line contains the name (label) of a sion and partial man-
sample:
tel
Stade de France
Twickenham Beginning of file YANOM.ALL of the distribution
Arms Park
Millenium Put here any comments you want
Lansdowne road 171 3
Murrayfield var0 (YANOM.GEN)
var1 (YANOM.ANT)
var2 (YANOM.GEO)
Only the first six characters (those in 35.00000
44.00000
96.00000
147.00000
9.00000
28.00000
bold) will be used in the output files 38.00000
47.00000
88.00000
295.00000
20.00000
152.00000
One digit encoding for data file Two 56.00000
59.00000
318.00000
309.00000
161.00000
178.00000
digits encoding for data file 52.00000
65.00000
284.00000
339.00000
149.00000
158.00000
67.00000 348.00000 175.00000
30.00000 158.00000 7.00000
57.00000 253.00000 169.00000
A.4 Example of file for- 76.00000

80.00000
50.00000
258.00000
260.00000
235.00000
175.00000
184.00000
102.00000
mat for testing bi- 43.00000

65.00000
69.00000
253.00000
289.00000
301.00000
95.00000
172.00000
178.00000
ased dispersal 69.00000

52.00000
60.00000
321.00000
277.00000
270.00000
187.00000
98.00000
91.00000
68.00000 120.00000 7.00000
File exampsb.gen of the distribution. 69.00000
74.00000
507.00000
571.00000
253.00000
259.00000
63.00000 583.00000 276.00000
File exampsb.gen of the distribution. 61.00000 519.00000 232.00000
65.00000 455.00000 238.00000
5 8 64 88.00000 472.00000 330.00000
loc-3ks, loc-17, loc-3T7, loc-57, loc-72, loc-23, loc-9, loc-53 79.00000 487.00000 327.00000
Pop 34.00000 488.00000 244.00000
F,0, 4041 2627 3234 2828 2830 2122 2223 2022 46.00000 560.00000 250.00000
M,0, 3940 2929 3232 3333 2430 2022 2329 0 46.00000 570.00000 267.00000
M,0, 4040 2727 3232 2833 2731 0 2222 2020
F,0, 4059 2429 3434 3233 2430 1920 2325 2323
F,0, 4059 2429 3232 3233 2430 1920 2329 2023
F,0, 4040 2429 3232 3233 3030 0 0 2222
F,0, 4041 2829 3234 2931 2428 1919 2223 2023
M,0, 4059 2930 3434 3233 2425 1920 2531 2020
M,0, 4040 2429 3234 3233 0 0 2323 2023
Pop
F,0, 4041 2629 3334 3031 2428 2021 2326 2021
F,0, 4040 2735 3434 2833 2528 2023 2326 2020
M,0, 4040 2929 3333 2931 2424 2022 2223 2020
M,0, 4040 2629 3334 3030 2425 2020 2326 2121
F,0, 3940 2929 3234 2932 2425 1919 2326 2023
M,0, 4040 2729 3333 2931 2430 0 2323 2023
M,0, 4040 2626 3434 2931 2830 1921 2326 2021
F,0, 4040 2629 3334 3031 2425 2020 2223 2121
F,0, 4041 2727 3234 3232 2428 2021 2326 2123
M,0, 4040 2729 3233 3031 2430 2022 2223 1920
Pop
F,0, 4040 2729 3234 2933 2425 1919 2223 2023
M,0, 4040 2727 3434 2929 2425 1919 2223 1821
M,0, 4041 2729 3434 0 2528 1922 2123 2323
M,0, 4040 2727 3234 2933 2527 1919 2326 1821
F,0, 4040 2429 3333 2929 2324 2223 2323 1818
M,0, 4040 2929 3232 2933 2427 1919 0 2121
M,0, 4040 2929 3434 2933 2525 1919 2326 2121
M,0, 4041 2727 3434 2932 2528 1919 2326 2121
M,0, 4041 2929 3434 3333 2528 1919 2226 2121
Pop
F,0, 4060 3031 1734 3134 2830 2224 2428 2324
F,0, 4157 2830 1734 2834 1332 2124 2330 1919
M,0, 4040 2335 3333 3032 2328 2224 2527 1922
M,0, 4041 2331 3234 3034 2532 2324 2527 1922
M,0, 4041 2832 3334 3034 2832 2224 2730 1919
F,0, 4158 2635 1732 3131 1323 2225 2426 1920
F,0, 4040 2729 3034 3131 2328 2225 2527 1919
F,0, 4040 2735 3334 3032 2323 2222 2525 1922
F,0, 4054 2628 1732 2929 2931 2023 2526 1920
M,0, 4041 2428 3434 2734 2532 2425 2325 1922
Pop
F,0, 4040 2629 3430 2830 2931 2323 2324 2222
M,0, 4041 2832 3434 2530 2932 2024 2529 1920
M,0, 4040 2328 3432 3031 2929 2020 2426 1920
M,0, 4041 2628 3034 2930 2929 2023 2426 2022
M,0, 4052 2731 3428 2530 3030 2024 2323 1920
M,0, 4141 2832 3434 2530 2932 2024 2323 1920
M,0, 4052 2832 3428 3134 3030 2022 2323 1920
F,0, 4360 2627 1729 2930 1213 1924 2426 1921
F,0, 4153 3332 3428 2534 2932 2224 2127 1920
F,0, 4040 2832 3433 2829 2929 2425 2126 1920
M,0, 4040 2826 3034 3930 2929 2023 2424 1922
Appendix B
Files of results from the

input file in Appendix 1.
B.1 DIPLOID.OUT
****************************************************************************************************************************
* The following results were generated the 09.08.2001 at 13:44:36 with \textsc{Fstat} for windows, V2.9.3 from file diploid2.dat. *
****************************************************************************************************************************
Stade Twicke Arms P Millen Lansdo Murray All_W All_UW
Locus: loc-1
N 8 8 5 7 9 7
p: 4 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000
Locus: loc-2
N 8 8 5 7 9 7
p: 3 0.063 0.250 0.000 0.000 0.056 0.000 0.068 0.061
p: 4 0.938 0.750 1.000 1.000 0.944 1.000 0.932 0.939
Locus: loc-3
N 5 8 5 7 9 7
p: 2 0.100 0.000 0.000 0.000 0.000 0.000 0.012 0.017
p: 3 0.400 0.313 0.400 0.357 0.167 0.143 0.280 0.297
p: 4 0.500 0.688 0.600 0.643 0.833 0.857 0.707 0.687
Locus: loc-4
N 8 8 5 7 8 7
p: 1 0.000 0.000 0.200 0.000 0.063 0.071 0.047 0.056
p: 2 0.000 0.188 0.200 0.000 0.063 0.143 0.093 0.099
p: 3 0.688 0.375 0.200 0.286 0.438 0.357 0.407 0.390
p: 4 0.313 0.438 0.400 0.714 0.438 0.429 0.453 0.455
Locus: loc-5
N 8 8 5 7 9 7
p: 4 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000
************************************************
Gene diversity per locus and population :

loc-1 0.000 0.000 0.000 0.000 0.000 0.000
loc-2 0.125 0.411 0.000 0.000 0.111 0.000
loc-3 0.600 0.482 0.550 0.476 0.292 0.262
loc-4 0.446 0.688 0.800 0.452 0.643 0.714
loc-5 0.000 0.000 0.000 0.000 0.000 0.000
************************************************
number of alleles sampled :

loc-1 1 1 1 1 1 1 1
loc-2 2 2 1 1 2 1 2
loc-3 3 2 2 2 2 2 3
loc-4 2 3 4 2 4 4 4
loc-5 1 1 1 1 1 1 1
************************************************
Allelic Richness per locus and population

based on min. sample size of: 5 diploid individuals.
loc-1 1.000 1.000 1.000 1.000 1.000 1.000 1.000
loc-2 1.625 1.992 1.000 1.000 1.556 1.000 1.526
loc-3 3.000 1.999 2.000 2.000 1.931 1.934 2.093
loc-4 1.999 2.964 4.000 1.999 3.250 3.648 3.035
loc-5 1.000 1.000 1.000 1.000 1.000 1.000 1.000
************************************************
Fis Per population :

loc-1 NA NA NA NA NA NA
loc-2 0.000 0.391 NA NA 0.000 NA
45
APPENDIX B. FILES OF RESULTS FROM THE INPUT FILE IN APPENDIX 1.46
loc-3 -0.667 0.741 0.273 -0.500 -0.143 -0.091

loc-4 -0.400 0.273 0.000 0.368 -0.167 0.000
All -0.494 0.446 0.111 -0.077 -0.142 -0.024
************************************************
P-value for Fis within samples.

based on : 3000 randomisations.
Indicative adjusted nominal level (5\%) for one table is : 0.00167
Proportion of randomisations that gave a LARGER Fis than the observed:

loc-2 1.0000 0.3950 NA NA 1.0000 NA
loc-3 1.0000 0.0807 0.6300 1.0000 1.0000 1.0000
loc-4 1.0000 0.2223 0.7030 0.4500 0.8533 0.6703
All 1.0000 0.0137 0.4757 0.8217 0.8710 0.6957
Proportion of randomisations that gave a SMALLER Fis than the observed:

loc-2 1.0000 0.9880 NA NA 1.0000 NA
loc-3 0.1233 1.0000 0.9547 0.3193 0.8310 0.9230
loc-4 0.3977 0.9233 0.7657 0.9817 0.4297 0.6733
All 0.0503 0.9953 0.8163 0.6247 0.3630 0.6287
************************************************
Nei’s estimation of heterozygosity
LocName Ho Hs Ht Dst Dst’ Ht’ Gst Gst’ $G_{IS}$

loc-1 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
loc-2 0.081 0.109 0.117 0.008 0.009 0.118 0.065 0.077 0.258
loc-3 0.476 0.443 0.445 0.002 0.002 0.446 0.004 0.005 -0.074
loc-4 0.613 0.623 0.635 0.013 0.015 0.638 0.020 0.024 0.016
loc-5 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Overall 0.234 0.235 0.239 0.004 0.005 0.240 0.018 0.022 0.005
************************************************
Weir & Cockerham (1984) estimation of Fit (CapF), Fst (theta) and Fis (smallF).
relat is Relatedness estimated following Queller & Goodnight (1989)
relatc is relatedness inbreeding corrected following Pamilo (1984, 1985)
sig_a, sig_b and sig_w are the component of variance
among samples, among individuals within samples and within individuals respectively.
For locus : loc-2

Allele Capf Theta Smallf Relat Relatc Sig_a Sig_b Sig_w
3 0.303 0.069 0.251 0.107
4 0.303 0.069 0.251 0.107
All 0.303 0.069 0.251 0.107 -0.668 0.009 0.030 0.091
For locus : loc-3

2 0.006 0.037 -0.032 0.073
3 -0.017 -0.010 -0.006 -0.021
4 -0.045 0.017 -0.063 0.036
All -0.030 0.005 -0.035 0.009 0.067 0.002 -0.015 0.439
For locus : loc-4

1 -0.030 0.046 -0.080 0.096
2 0.186 0.011 0.177 0.019
3 0.007 0.048 -0.043 0.095
4 0.027 0.002 0.025 0.004
All 0.037 0.024 0.013 0.047 -0.026 0.015 0.008 0.605
************************************************
Over all loci

Capf Theta Smallf Relat Relatc Sig_a Sig_b Sig_w
0.042 0.022 0.020 0.043 -9.990 0.026 0.023 1.135
************************************************
Rst Over all samples estimated following Rousset (1996) and Goodman (1997)
Rst siga sigb sigw amean astdev

loc-2 0.069 0.0 0.0 0.0 3.93 0.2521
loc-3 0.042 0.0 -0.0 0.3 3.70 0.4861
loc-4 0.016 0.0 0.2 0.4 3.27 0.8130
Rst over loci Weighted Goodman Unweighted

Rst: 0.0260 0.0426 0.0425
************************************************
************************************************
Jackknifing over populations.
For locus : loc-2

Capf Theta Smallf Relat
Total 0.461 0.117 0.357 0.207 Means
0.303 0.100 0.226 0.162 Std. Err.
For locus : loc-3

Total -0.032 0.003 -0.025 -0.012 Means
0.222 0.046 0.264 0.108 Std. Err.
For locus : loc-4

Total 0.035 0.023 0.014 0.046 Means
0.098 0.058 0.106 0.111 Std. Err.
************************************************
Jackknifing over loci.

total 0.031 0.021 0.010 0.041 Means
0.046 0.010 0.038 0.018 Std. Err.
************************************************
Bootstrapping over Loci.
95\% Confidence Interval.
CapF theta Smallf Relat

-0.030 0.005 -0.035 0.009
0.303 0.069 0.251 0.107
99\% Confidence Interval.
CapF theta Smallf Relat

-0.030 0.005 -0.035 0.009
0.303 0.069 0.251 0.107
************************************************
Randomising alleles within samples.

Testing for Hardy-Weinberg within samples.
Statistic used to classified tables is smallf (Fis).
based on : 5000 randomisations.
loc-2 Prop rand. larger than observed in [ 0.39600 0.01320]

All Loci Prop rand. larger than observed in [ 0.41180 0.39980]
************************************************
Randomising alleles overall samples
Test based on : 5000 randomisations.
Testing for Hardy-Weinberg overall using the statistic CapF (Fit)

************************************************
Randomising genotypes among samples.
test based on : 5000 randomisations.
testing for population differentiation,

NOT ASSUMING RANDOM MATING within samples.
Statistic used is the log-likelihood G (Goudet et al, 1996)

*******************************************************
P-value for genotypic disequilibrium
based on 6000 permutations.
Adjusted P-value for 5\% nominal level is : 0.000833
Adjusted P-value for 1\% nominal level is : 0.000167
Stade Twicke Arms P Millen Lansdo Murray All

loc-1 X loc-2 NA NA NA NA NA NA NA
loc-2 X loc-3 1.00000 0.77000 NA NA 1.00000 NA 0.81650
loc-2 X loc-4 0.36633 1.00000 NA NA 1.00000 NA 0.97733
loc-3 X loc-4 1.00000 0.42883 0.34117 0.33250 1.00000 0.51867 0.21267
B.2 Files obtained from the ”FST per pair” option

DIPLOID.FST
0.0000 0.0207 0.0665 0.1101 0.0704 0.1068

0.0207 0.0000 -0.0390 0.0119 -0.0185 -0.0131
0.0665 -0.0390 0.0000 -0.0006 0.0147 -0.0104
0.1101 0.0119 -0.0006 0.0000 0.0342 0.0378
0.0704 -0.0185 0.0147 0.0342 0.0000 -0.0525
0.1068 -0.0131 -0.0104 0.0378 -0.0525 0.0000
DIPLOID.MAT
1.3980
1.3177 1.4438
1.1724 1.2953 1.0982
1.1452 1.2730 1.1690 1.0243
1.1648 1.2850 1.1157 0.9898 0.9666
0.0290
0.0876 -0.0563
0.1291 0.0155 -0.0006
0.0806 -0.0235 0.0171 0.0351
0.1245 -0.0168 -0.0116 0.0374 -0.0508
B.3 DIPLOID.X2
Observed and expected genotype frequencies
Observed for locus : loc-1

0404 8 8 5 7 9 7
Expected for locus : loc-1

0404 8.00 8.00 5.00 7.00 9.00 7.00

0303 0 1 0 0 0 0
0304 1 2 0 0 1 0
0404 7 5 5 7 8 7

0303 0.00 0.40 0.00 0.00 0.00 0.00
0304 1.00 3.20 0.00 0.00 1.00 0.00
0404 7.00 4.40 5.00 7.00 8.00 7.00

0202 0 0 0 0 0 0
0203 0 0 0 0 0 0
0204 1 0 0 0 0 0
0303 0 2 1 0 0 0
0304 4 1 2 5 3 2
0404 0 5 2 2 6 5

0202 0.00 0.00 0.00 0.00 0.00 0.00
0203 0.44 0.00 0.00 0.00 0.00 0.00
0204 0.56 0.00 0.00 0.00 0.00 0.00
0303 0.67 0.67 0.67 0.77 0.18 0.08
0304 2.22 3.67 2.67 3.46 2.65 1.85
0404 1.11 3.67 1.67 2.77 6.18 5.08

0101 0 0 0 0 0 0
0102 0 0 2 0 1 0
0103 0 0 0 0 0 0
0104 0 0 0 0 0 1
0202 0 1 0 0 0 0
0203 0 1 0 0 0 1
0204 0 0 0 0 0 1
0303 3 1 0 1 1 1
0304 5 3 2 2 5 2
0404 0 2 1 4 1 1

0101 0.00 0.00 0.11 0.00 0.00 0.00
0102 0.00 0.00 0.44 0.00 0.07 0.15
0103 0.00 0.00 0.44 0.00 0.47 0.38
0104 0.00 0.00 0.89 0.00 0.47 0.46
0202 0.00 0.20 0.11 0.00 0.00 0.08
0203 0.00 1.20 0.44 0.00 0.47 0.77
0204 0.00 1.40 0.89 0.00 0.47 0.92
0303 3.67 1.00 0.11 0.46 1.40 0.77
0304 3.67 2.80 0.89 3.08 3.27 2.31
0404 0.67 1.40 0.67 3.46 1.40 1.15

0404 8 8 5 7 9 7

0404 8.00 8.00 5.00 7.00 9.00 7.00
B.4 Files obtained from the ”testing pairwise pop

differentiation” option
Twicke Arms P Millen Lansdo Murray
Stade 0.14467 0.02733 0.03467 0.21000 0.03733
Twicke 0.31867 0.18667 0.40400 0.05733
Arms P 0.27800 0.61067 0.53467
Millen 0.43733 0.29000
Lansdo 0.95400
P-values obtained after : 1500 permutations

Indicative adjusted nominal level (5\%) for multiple comparisons is : 0.003333
Twicke Arms P Millen Lansdo Murray

Stade NS NS NS NS NS
Twicke NS NS NS NS
Arms P NS NS NS
Millen NS NS
Lansdo NS
B.5 Example of result file from the comp. Among

group of samples tab sheet file diploid-test.out
of the distribution
file diploid-test.out of the distribution
****************************************************************************************************************************
* The following results were generated the 03.08.2001 at 16:26:37 with \textsc{Fstat} for windows, V2.9.3 from file diploid2.dat. *
****************************************************************************************************************************
Samples of group 1 (G1) : pop1, pop2, pop3,
Allelic richness: 1.772 Ho: 0.241 Hs: 0.273 Fis: 0.117 Fst: 0.013 Rel: 0.024 Relc: -0.264
Samples of group 2 (G2) : pop4, pop5, pop6,
Allelic richness: 1.621 Ho: 0.214 Hs: 0.197 Fis: -0.083 Fst: 0.005 Rel: 0.011 Relc: 0.153
The two-sided P-values obtained after 5000 permutations are :

Allelic Richness: 0.19440
Ho: 0.80540
Hs: 0.10380
Fis: 0.70200
Fst: 0.96200
Rel: 0.96200
Relc: 0.70200
B.6 Example of result file from the biased disper-

sal menu
File exampsb.res of the distribution
results are for data stored in filename: examp

The tests are two sided and based on 1000 randomisations.
Fis for M: -0.0280; Fst for M: 0.2109
Fis for F: -0.0045; Fst for F: 0.0769
Fis overall: -0.0069; Fst overall: 0.1377
NB M: 27; mean assignment: 1.51030; var assignment: 4.11501
NB F: 22; mean assignment: -1.85355; var assignment: 16.25338
p_value for assignment Ttest: 0.0020
p_value for variance assignment test: 0.0060
p_value for Fst test: 0.0020
p_value for Fis test: 0.7240
B.7 Example of result file from the multiple re-

gression and partial mantel menu
File YANOM.RES of the distribution
Put here any comments you want data from the file :
E:\fstatdev\Fstat2.9.3\data\multiple regression\YANOM.ALL
P-values are given after 2000 randomizations.
The total sum of square for variable var0 (YANOM.GEN) is :30312.4219
Correlation (Partial if # expl. variables > 1), Coefficient (Beta) and Sum of squares (SS) for the observed data:
Variable (Partial) Corr. Beta SS
------------------------------------------------------------
var1 (YANOM.ANT) 0.299551 0.029754 2719.9512
------------------------------------------------------------
Error sum of squares: 27592.4707
Percent of the variance explained by the model (R_squared): 8.97
Percent. point for absolute regression coefficients (2-sided) P(Beta)

and percent. point for extra sums of squares P(SS)
Variable P(Beta) P(SS)

------------------------------
var1 (YANOM.ANT) 0.05 0.05
------------------------------
Percent point for the error sum of square 0.05
Bibliography
F. Balloux. Easypop (version 1.7): a computer program for population genet-

ics simulations. Journal of Heredity, 2001. URL http://www.unil.ch/izea/
softwares/easypop.html.
F. Balloux, H. Brünner, N. Lugon-Moulin, J. Hausser, and J. Goudet. Mi-
crosatellites can be misleading: an empirical and simulation study. Evolution,
54:1414–1422, 2000. URL http://www.unil.ch/popgen/research/reprints/
ballouxetal_evolution_2000.pdf.
F. Balloux and J. Goudet. Statistical properties of population differentiation es-
timators under stepwise mutation in a finite island model. Molecular Ecology,
balloux&goudet_mec_2002.pdf.
R. Chakraborty and H. Danker-Hopfe. Statistical Methods in Biological and Med-
ical Sciences, chapter Analysis of population structure: A comparative study of
different estimators of Wright’s fixation indices. Elsevier Science Publishers, 1991.
M. Chapuisat, J. Goudet, and L. Keller. Microsatellites reveal high population
viscosity and limited dispersal in the ant formica paralugubris. Evolution,
chapuisatetal_evolution_1997.pdf.
C.C. Cockerham. Variance of gene frequencies. Evolution, 23:72–84, 1969.
C.C. Cockerham. Analysis of gene frequencies. Genetics, 74:679–700, 1973.
C.C. Cockerham and B.S. Weir. Estimation of gene-flow from F -statistics. Evolu-
tion, 47:855–863, 1993.
A. El Mousadik and R.J. Petit. High level of genetic differentiation for allelic
richness among populations of the argan tree [argania spinosa (l.) skeels] endemic
to morocco. Theoretical and Applied Genetics, 92:832–839, 1996.
L. Excoffier. Handbook of statistical genetics, chapter Analysis of population sub-
division. Wiley and Sons, Ltd, 2001.
L. Favre, F. Balloux, J. Goudet, and N. Perrin. Female-biased dispersal
in the monogamous mammal crocidura russula: evidence from field data
and microsatellite patterns. Proceeding of the Royal Society of London B,
favreetal_prsb_1997.pdf.
R. Fisher. Statistical Methods for Research Workers. Oliver & Boyd, Edinburgh,
12 edition, 1954.
50
BIBLIOGRAPHY 51
S.J. Goodman. Rst calc: a collection of computer programs for calculating esti-
mates of genetic differentition from microsatellite data and a determining their
significance. Molecular Ecology, 6:881–885, 1997. URL http://helios.bto.ed.
ac.uk/evolgen/rst/rst.html.
J. Goudet. The genetics of geographically structured populations. PhD thesis, Uni-
versity of Wales at bangor, 1993. URL http://www.unil.ch/popgen/research/
reprints/.
J. Goudet. Fstat (vers.1.2): a computer program to calculate F -statistics. Jour-
nal of Heredity, 86:485–486, 1995. URL http://www.unil.ch/izea/softwares/
fstat.html.
J. Goudet. An improved procedure for testing key innovations. American Natural-
ist, 53:549–555, 1999. URL http://www.unil.ch/popgen/research/reprints/
goudet_amnat_1999.pdf.
J. Goudet, T. De Meeüs, A.J. Day, and C.J. Gliddon. Genetics and Evolution
of Aquatic Organisms, chapter The different levels of population structuring of
dogwhelks, Nucella lapillus, along the south Devon coast. Chapman and Hall,
London, 1994. URL http://www.unil.ch/popgen/research/reprints/.
J. Goudet, N. Perrin, and P. Waser. Tests for sex-biased dispersal using bi-parentally
inherited genetic markers. Molecular Ecology, 11:1103–1114, 2002. URL http:
//www.unil.ch/popgen/research/reprints/goudetetal_mec_2002.pdf.
J. Goudet, M. Raymond, T. Demeeus, and F. Rousset. Testing differentiation in
diploid populations. Genetics, 144:1933–1940, 1996. URL http://www.unil.ch/
popgen/research/reprints/goudetetal_genetics_1996.pdf.
W.D. Hamilton. Man and Beast: Comparative Social Behavior, chapter Selection of
selfish and altruistic behaviour in some extreme models, pages 57–91. Eisenberg
and Dillon, washington, DC, 1971.
D.L. Hartl and A.G. Clarck. Principles of Population Genetics. Sinauer Associates,
third edition, 1997.
P.W. Hedrick. Genetics of Population. Jones and Bartlett, second edition, 2000.
K. E. Holsinger, P. O. Lewis, and D. K. Dey. A bayesian method for analysis
of genetic population structure with dominant marker data. Molecular Ecology,
11:1157–1164, 2002. URL http://darwin.eeb.uconn.edu/hickory/hickory.
html.
S.H. Hurlbert. The nonconcept of species diversity: a critique and alternative
parameters. Ecology, 52:577–586, 1971.
P. L’Ecuyer. Efficient and portable random number generators. Communications
of the ACM, 31:147–157, 1988.
R.C. Lewontin and K. Kojima. The evolutionary dynamics of complex polymor-
phism. Evolution, 14:458–472, 1960.
B.J.F. Manly. The Statistics of Natural Selection. Chapman and Hall, 1985.
B.J.F. Manly. Randomization and Monte Carlo methods in biology. Chapman et
Hall., second edition, 1997.
M. Nei. Analysis of gene diversity in subdivided populations. Proceedings of the
National Academy of Sciences USA, 70:3321–3323, 1973.
BIBLIOGRAPHY 52
M. Nei. Molecular Population Genetics and Evolution. Elsevier, 1975. URL http:
//www.bio.psu.edu/People/Faculty/Nei/Lab/BOOK.pdf.
M. Nei. Molecular Evolutionary Genetics. Columbia University Press, 1987.
M. Nei and R.K. Chesser. Estimation of fixation indices and gene diversities. Annals
of Human Genetics, 47:253–259, 1983.
R. D. M. Page. Treeview: An application to display phylogenetic trees on personal

computers. Computer Applications in the Biosciences, 12:357–358, 1996. URL
http://taxonomy.zoology.gla.ac.uk/rod/treeview.html.
P. Pamilo. Genotypic correlation and regression in social groups: multiple alleles,
multiple loci and subdivided populations. Genetics, 107:307–320, 1984.
P. Pamilo. Effect of inbreeding on genetic relatedness. Hereditas, 103:195–200, 1985.

E. Petit, F. Balloux, and J. Goudet. Sex biased dispersal in a migratory
bat: a characterization using sex-specific demographic parameters. Evolution,
petitetal_evolution_2001.pdf.
R.J. Petit, A. El Mousadik, and O. Pons. Identifying populations for conservation
on the basis of genetic markers. Conservation Biology, 12:844–855, 1998.
J.K. Pritchard, M. Stephens, and P. Donnelly. Inference of population structure
using multilocus genotype data. Genetics, 155:945–959, 2000. URL http://
pritch.bsd.uchicago.edu/.
D.C. Queller and K.F. Goodnight. Estimating relatedness using genetic markers.
Evolution, 42:258–275, 1989.
N. Raufaste and F. Rousset. Are partial mantel tests adequate? Evolution, 55:
1703–1705, 2001.
M. Raymond and F. Rousset. An exact test for population differentiation. Evolution,
49:1280–1283, 1995a.
M. Raymond and F. Rousset. Genepop (version 1.2): population genetics software
for exact tests and oecumenicism. Journal of Heredity, 86:248–249, 1995b. URL
ftp://ftp.cefe.cnrs-mop.fr/genepop/.
J. Reynolds, B.S. Weir, and C.C. Cockerham. Estimation of the coancestry coeffi-
cient: Basis for a short-term genetic distance. Genetics, 105:767–779, 1983.
W.R. Rice. Analysing tables of statistical tests. Evolution, 43:223–225, 1989.
F. Rousset. Equilibrium values of measures of population subdivision for stepwise

mutation processes. Genetics, 142:1357–1362, 1996.
F. Rousset. Genetic differentiation and estimation of gene flow from F -statistics
under isolation by distance. Genetics, 145:1219–1228, 1997.
S. Schneider, D. Roessli, and L. Excoffier. Arlequin: A software for population genet-
ics data analysis. Genetics and Biometry Lab, Dept. of Anthropology, University
of Geneva, 2.000 edition, 2000. URL http://lgb.unige.ch/arlequin/.
M. Slatkin. A measure of population subdivision based on microsatellite allele
frequency. Genetics, 139:457–462, 1995.
BIBLIOGRAPHY 53
M. Slatkin and N.H. Barton. A comparison of three methods for estimating average
levels of gene flow. Evolution, 43:1349–1368, 1989.
S. Trouvé, L. Degen, F. Renaud, and J. Goudet. Population structure of the fresh-
water snail lymnaea truncatula: the importance and consequences of selfing.
Evolution, 57:In Press, 2003. URL http://www.unil.ch/popgen/research/
reprints/trouveetal_evolution_2003_pp.pdf.
J. Wang. An estimator for pairwise relatedness using molecular markers. Genetics,

160:1203–1215, 2002.
B.S. Weir. Inferences about linkage disequilibrium. Biometrics, 35:235–254, 1979.
B.S. Weir. Genetic data analysis II. Sinauer, second edition, 1996.
B.S. Weir and C.C. Cockerham. Estimating F -statistics for the analysis of popula-
tion structure. Evolution, 38:1358–1370, 1984.
M.C. Whitlock and D. McCauley. Indirect measures of gene flow and migration:
FST 6= 1/(4N m + 1). Heredity, 82:117–125, 1999.
S. Wright. Systems of mating. Genetics, 6:111–178, 1921.
S. Wright. Evolution and the genetics of populations. II. The theory of gene fre-
quencies, volume 2. University of Chicago Press, 1969.
C. Zapata. The D0 measure of overall gametic disequilibrium between pairs of
multiallelic loci. Evolution, 54:1809–1812, 2000.
C. Zapata, C. Carollo, and Rodriguez S. Sampling variance and distribution of the
D0 measure of overall gametic disequilibrium between multiallelic loci. Annals of
Human Genetics, 65:395–406, 2001.
D. Zaykin, L. Zhivotovsky, and B.S. Weir. Exact tests for association between alleles
at arbitrary numbers of loci. Genetica, 96:169–178, 1995.

Fstat 294

Uploaded by

Copyright:

Available Formats

Fstat 294

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Fstat 294

Uploaded by

Copyright:

Available Formats

FSTAT (version 2.9.

4), a program (for Windows 95

21st November 2003

2 Running the program 8

4 Choose options menu 11

5 Per locus and sample statistics 13

7.3.1 HW tests per locus and sample . . . . . . . . . . . . . . . . . 21

10 Comp. Among groups of samples 28

15 Special use of Fstat 41

A Examples of input files 43

B Files of results from the input file in Appendix 1. 45

For evolution to occur, there must be polymorphism on which selection can

Do not hesitate to email me if you have suggestions for improvements, discover

1.1 What the program FSTAT does

• Estimation of FST (θ) per pair of samples.

1.1.1 What’s new in version 2.9.4?

1.2 What the program FSTAT does NOT do

Running the program

2.1 Creating an input file

2.2 LIMITS IN FSTAT

• Maximum number of individuals : 200’000

This is the only accessible menu when you start Fstat.

3.1 Open submenu

3.2 Save Default Options subMenu

3.3 Exit subMenu

Choose options menu

4.1 Label file for pops sub-menu

4.2 Loci to use sub-menu

4.3 Samples to use sub-menu

Per locus and sample

5.1 Allele frequencies

5.2 Genotypic frequencies

where Ai Ai represents homozygote genotypes, Ai Aj represents heterozygotes, n

5.3 Number of alleles

5.4 Allelic Richness

The principle is to estimate the expected number of alleles in a sub-sample of 2n

5.5 Gene diversities

(see [Nei, 1987, eq. 7.39 p. 164]).

6.1 Nei’s estimators of F-statistics

where Pkii represents the proportion of homozygote i in sample k and np the

where ñ = P np and p2i = p2ki /np

• the overall gene diversity

• the amount of gene diversity among samples DST = HT − HS

6.2 Weir & Cockerham estimators of F-statistics

and the within individual component

6.2.1 Jackknife variance and Bootstrap confidence interval

where l represents the number

the average of the individual Loci RST .

6.4 FST per pair

For HW within samples

For population differentiation

where l indexes loci.

Test overall loci

7.2 Global tests

7.2.2 Hardy Weinberg Overall samples

7.2.3 Population Differentiation NOT assuming Hardy Wein-

7.2.4 Population Differentiation assuming Hardy Weinberg

7.3 Tests per sample or pair of samples

7.3.2 Pairwise tests of differentiation