statistics books

I'd originally hoped to do a comprehensive survey of statistics books for this

tutorial. For various reasons (that I'll describe below), this didn't quite happen
as planned.
I first asked my friendly Prentice Hall rep. to send me any books intended for
non-statisticians, titles like "statistics for social sciences" etc. For the next few
weeks, almost every day, a parcel arrived! When the parcels stopped coming
and my bookshelves were bending under the weight, I decided that one
publisher was enough!
A complete list of these and other books I've looked at is at the end of this
section, but the majority are from Prentice Hall! For general statistics books
most publishers will have a similar range, so treat this as a 'random' sample.
The most interesting books seem to be the one-offs, special books not written in
the general hackneyed style of statistics books. If you find any that you find
particularly good for your own use or for students' use let me know and I'll add
them to the tutorial web site.

classics
The classic book about the (accidental and deliberate) abuse of statistics is:
¥ How to Lie with Statistics, Darrell Huff. Penguin Books, London. 1991.
(first published as How to Lie with Statistics, Darrell Huff and Irving
Geis, Victor Gollancz, 1954)
This deals with both numerical and graphical ways in which statistics confuse,
mislead and generally tell lies! It's also worth getting for Mel Calman's cartoons.
Of course, if you are producing graphs and want to do it well, then there is no
better place to go than Tufte:
¥ The Visual Display of Quantitative Information, E.R. Tufte. Graphics
Press, New York. 1983. (don't miss the 4,340Êlb chicken on page 73)
¥ Envisioning Information, E.R. Tufte. Graphics Press, New York. 1990.

Avoiding Damned Lies Ð Understanding Statistical Ideas Alan Dix www.meandeviation.com 19

 stats with no maths?
I was recently talking with a colleague who teaches statistics in a business
school, who told me of a book which "teaches statistics without using any
mathematics".
¥ Statistics without Tears, Derek Rowntree. Penguin Books.
Unfortunately, I haven't got myself a copy yet at the time I'm writing these
notes, but will do so soon and update the web bibliography accordingly.

who's afraid of ...

For many an anxiety about mathematics starts in early, in the first years of
school. There are various reasons:
¥ the progressive nature of the subject means that if once you loose track
you've had it for ever
¥ the unforgiving precision of arithmetic leaves no room for getting better
gradually
¥ the ability to follow rules and get the right answer without clear
understanding means teachers may not realise there is a problem until too late
¥ and last, but by no means least, most teachers are themselves terrified of
mathematics
So, perhaps it is at school rather than at an international academic conference
that real effort is needed!
Even if we were to be able to put this right overnight, those who have left their
school years behind can't simply forget this long established anxiety.
Among the books I've been looking at recently is:
¥ Statistics for the Terrified, Gerald Kranzler and Janet Moursund.
Prentice Hall, Englewood Cliffs, NJ. 1995.
This starts from the assumption that the reader has just such a deep maths
anxiety. The writer teaches statistics on counselling and education programmes
and draws on both, even including an appendix on "Overcoming Math Anxiety"!
The book is written in short punchy chapters, but does not patronise the reader
and does include the statistical formulae (even if you're afraid of maths, you still
need it!). I noticed few minor misunderstandings in the statistical details (page
15 says the wrong variance formulae minimises round off error and page 25
conflates two very different kinds of cumulative graph), but overall it looks
good. In particular, chapter 7 gives a very good introduction to the principles of
hypothesis testing and on page 10 there is an example of the difference between
mean and median just like the one I give!

Avoiding Damned Lies Ð Understanding Statistical Ideas Alan Dix www.meandeviation.com 20

 for the bold ...
In fact, those strong at mathematics often also find statistics difficult. They do all
the formulae right, but miss out on the fact that the numbers all mean
something! Statistics requires a blend of mathematics and the real world which
is challenging and exciting. Quite like the conflicts and joys of working with
computers and people!
Although the fundamental understanding of statistical concepts such as
significance testing don't require deep mathematics (but do require deep
thought!), the initial calculation of the formulae for these often do. This is
especially important when new forms of statistical problem are encountered.
For example, if you toss 20 coins you can look up in a standard table of the
binomial distribution the probability, say, of getting 12 heads and 8 tails.
However, recall the two horse races. You stopped when either heads or tails got
to 10. What about the shorter column. How long do you expect that to be?
Don't worry if you can't answer that quickly, it took me half an hour to work out
and I won't bother to quote the formulae here. Suffice to say it isn't in the
standard tables.
To help answer questions like this there is an extensive mathematical theory of
statistics, some about particular kinds of statistical problem (for example time
series analysis) and some about very general statistical properties (for example
the central limit theorem that says nearly everything is approximately normal if
its made of enough small bits). You don't need to understand any of this (nor
70% of what is in most general statistics books!) for day-to-day statistics.
However, if you need to do something out of the ordinary, you need to get
someone who can apply this general mathematical battery!
The two books that I refer back to for basic principles on this are:
¥ Theoretical Statistics, D. R. Cox and D. V. Hinkley. Chapman and Hall,
London. 1974.
¥ Mathematical Statistics 4e, John E. Freund and Ronald E. Walpole
Prentice Hall, Englewood Cliffs, NJ. 1987.
Did you notice that little word I slipped in ... 'basic' Ð do beware of any
mathematics books entitled 'basic', 'introductory' or 'elementary' anything Ð they
are usually advanced textbooks. In contrast, many that say 'advanced' or
'higher' are high school level!

Avoiding Damned Lies Ð Understanding Statistical Ideas Alan Dix www.meandeviation.com 21

 general statistics textbooks
The general format of statistics books is rather similar ... a bit about graphs, a bit
about probability and bit about significance testing. But there are differences in:
¥ level of detail Ð general statistics textbooks I looked at varied from 230 to
1182 pages!
¥ pedagogy Ð some in almost programmed learning style, others adopting a
more 'how to do it' approach
¥ required prior knowledge Ð engineers are expected to know quite
sophisticated mathematics before they start!
¥ examples Ð all the 'statistics for ...' textbooks tuned their examples to the
relevant target audience.
¥ choice of subject matter Ð although broadly similar in content different
domains employ specialised statistical tests or graphical displays (some I'd
never seen before!), for example, both engineering and business books use
'control charts' to check that processes are running within specified limits,
also some books spend more time on experimental method others on
statistical modelling
¥ use of computers Ð many have some exercises targeted at computers uses,
most commonly using M I N I T A B , but others rely entirely on hand
computation; perhaps most surprising is only one book mentions use of
spreadsheets (Excel) which for many people would be the obvious tool for
small data sets.

cover story
Several hardback books I looked at put useful tables of formulae or statistical
values on the inside front and back covers. Very useful for a book you buy
yourself, but of course totally useless if the book is purchased by a library and
has various labels stuck on its covers. Oh when will publishers realise that
libraries need somewhere to put their labels!

finding things
I wondered how well various statistics books functioned as how-to-do-it guides,
either as a reference or as a revision aid for students. As a test case I tried to look
up regression in each assuming I wanted to work out the line of best fit through
some data points. To make things easy I assumed I knew enough that it was the
'regression' equation I was after Ð I didn't try looking up line (just as well, the
few I looked at the word didn't appear).
First thing. In none of the books I looked at was there any indication in the
index of the defining location within the book. The closest I found was in.

Avoiding Damned Lies Ð Understanding Statistical Ideas Alan Dix www.meandeviation.com 22

 ¥ Elementary Statistics 2e, Neil Weiss. Addison Wesley, Reading, MA.
1993.
Although the ordinary index doesn't show defining references, there is a handy
'procedure index' on the inside back cover (yes the cover!). This has part
labelled 'regression' Ð I quote:
Slope of the population regression line
Confidence intervals, 639
Hypothesis tests, 636
Oops, no entry for the actual calculation of the slope!
In this and nearly all cases the best course of action was to find the relevant
chapter through the table of contents and to skim the chapter until I found the
formulae.
In fact, this book also has a pull-out reference card which includes the regression
formula, but again, alas, no reference back to the text.
(Incidentally, I have had this book for several years since an Addison Wesley
rep. called round and assured me she had a really good book on statistics.
Unfortunately, I found its rather noisy two colour scheme and programmed
learning approach almost impossible to read. However, this is very much a
matter of taste and looking at it again during this review process I found its
mini-biographies and motivating case studies very engaging.)
Full marks to:
¥ The Essence of Statistics for Business 2e, Michael C. Fleming and Joseph
G. Nellis. Prentice Hall, Hemel Hempstead 1996.
In the index I found:
regression equation 183, 197
Joy of joys, page 183 has the actual equation:

Formulae for regression coefficients:

ÊÊ∑Xi Y i ÊÐÊn X Y
b =
2
ÊÊ∑Xi ÊÐÊn X 2 ÊÊ

a = Y Ðb X
where Y and X are the arithmetic means of the Yi and Xi values
respectively

Furthermore, page 197 is the chapter-end summary of 'Key learning points'

which lists all the major terms and equations in the chapter.
Second prize goes to:
¥ Statistics for the Social Sciences Victoria L. Mantzopoulos. Prentice Hall,
Englewood Cliffs, NJ. 1995.

Avoiding Damned Lies Ð Understanding Statistical Ideas Alan Dix www.meandeviation.com 23

 This book has a lovely mini-dictionary/glossary in appendix B. Unfortunately
the glossary doesn't refer back to key pages ion the book! Never mind, I turn to
the index:
Regression, 350Ð51, 370
bi-variate, 64Ð67, 69
multiple, 67Ð69
I wisely ignore the two references late in the book (actually to a case study and
the mini-dictionary entry!) and plumb for 'bi-variate'. Bingo! The formulae for
the regression line and how to calculate the various coefficients in three grey
boxes.
Sadly this is the best there is. A more typical example was:
¥ Basic Statistics for the Social and Behavioural Sciences George M.
Diekhoff. Prentice Hall, Englewood Cliffs, NJ. 1996.
A nicely presented book with good use of margins and two colour to highlight
key definitions and equations. It has a glossary, but with no cross references to
the text! It also has 'Essential Equations' listed on inside front and back covers
(!!). On the inside back cover I find:

MAKING PREDICTIONS

s
Eq.11.1 Y' = a + bYX where b Y = r Y
 sX
a = Y Ð bY X

On the inside front cover I find the definition of ' X ' and of 's', and, let's say, I
remember that 'sY' and 'sY' are applications of the same formula as 's' with values
of Y and X respectively. But what is 'r'? A few lines above there is a definition
of 'rp', which has nothing to do with it! (I know this, does the typical reader?)
I turn to the index and after a few false starts find equation 11.2 which is the
formula for bY. (Incidentally 11.1 is just Y' = a + b YX.) This also has the ellusive
'r' but also an explanation:

b Y = the correlation between X and Y

At last I'm on to something. I look up 'correlation' in the index. This has some
main entries and lots of sub-entries. The main entries are no good and rather
than scanning all the subentries, I look back at the text around equation 11.2.
Just below, in the midst of the text I read 'multiply the resulting value by the
Pearson correlation between X and Y (r)' Ð at last!! I look up 'Pearson correlation'
and get there.
Note that this doesn't mean that Diekhoff is particularly bad, in fact overall it
looks pretty good (even mentions squared dollars!), it's just that books in general
tend to have pretty poor indices and statistics books are no exception.
Unfortunately this makes them pretty useless when you want to use them as a
how-to-do-it reference.

Avoiding Damned Lies Ð Understanding Statistical Ideas Alan Dix www.meandeviation.com 24

 tables
Most textbooks include some statistical tables in an appendix. For more detail
you may want to get a special book of statistical tables. The one I use is:
¥ Statistical Tables, F.D.J. Dunstan, A.B.J. Nix and J.F. Reynolds. R.N.D
Publications, Cardiff. ISBN 0-9506719-0-8. 1979.
Another popular book of tables is:
¥ Elementary Statistical Tables, Henry R. Neave. Allen & Unwin, London.
I don't know if this is a problem in other countries, but British students now
have great difficulty reading tables because the use of calculators has meant that
old logarithmic and trigonometric tables are no longer used in school. For
students' use I would recommend avoiding any books of tables that have
'interpolation columns' as these are particularly confusing (albeit useful).
Also, a word of warning, be very careful with:
¥ value → significance vs. significance → value. Some tables are listed by
significance value, often at 5%, 2.5%, 1% and 0.5%. You look up the
significance you are after, say 5%, and then the table tells you the 'critical
value', that is if your calculated value is bigger than this critical value it is 5%
significant. Other tables, notably tables of the Normal distribution and also
discrete distributions such as the binomial work the other way round. You
look up your value in the table and then it tells you the probability
(significance) of that value.
¥ probability or percentage. Tables may use probabilities (e.g. 0.05) or
percentages (e.g. 5%) and may also list values by the probability that a value
is less than a value, which may mean you have to look up the 0.95
(probability) or 95% (percentage) point.
¥ one- and two-tailed tables. First you need to know which you want Ð no
easy matter even when you know what the words mean. Then you have to
look it up... For asymmetric distributions (e.g. ℵ2) both tables may be given
(or combined in a single dual purpose table), for symmetric distributions one
or other may be given in the tables and you are expected to work out the
other (e.g. for Students' t distribution to find the two-tailed 5% point you
look up the 2.5% one-tailed value from the tables!). Worst of all is the F
distribution. Most frequently this is used in ANOVA tests and the tables are
hence given in one-tailed form. However, occasionally you need the two-
tailed version and then you have to work out a weird formulae including the
reciprocal of the value in the table (Just hope you never need that one!).
¥ Normal distribution. These are usually given in one-tailed form (although
sometimes two-tailed as well). But they have at least three different forms.
(i)ÊFirst you have tables which give the one-tailed significance value Ð that is
the probability that a value is greater than the given value). (ii)ÊOthers give
the probability that the value is less than the quoted value. (iii)ÊFinally some
give the probability of a value between zero and the given value. If you
looked up the value 1.65 in these tables you will find: (i)Ê0.05 (ii)Ê0.95
Avoiding Damned Lies Ð Understanding Statistical Ideas Alan Dix www.meandeviation.com 25
 (iii)Ê0.45. Note that the Normal table in 'Statistics for the Terrified' gives both
(iii) and (i) side by side.
Many statistics packages still expect you to look up the value in a table, so this,
possibly one of the most confusing aspects of statistical computation, is not
always eased by the use of computers.

book list
Here is the complete list of books that I've considered. A preponderance of
Prentice Hall titles for the reasons I gave earlier! Do tell me about other books
you find useful, especially those that address conceptual understanding.
¥ How to Lie with Statistics, Darrell Huff. Penguin Books, London. 1991.
(first published as How to Lie with Statistics, Darrell Huff and Irving
Geis, Victor Gollancz, 1954) (124 pages)
¥ I've been meaning to get my own copy of this for years and eventually did ready for
this tutorial. Well worth it!
¥ The Visual Display of Quantitative Information, E.R. Tufte. Graphics
Press, New York. 1983.
¥ The classic book of graphical representation of data. Don't miss the 4,340Êlb chicken on
page 73.
¥ Envisioning Information, E.R. Tufte. Graphics Press, New York. 1990.
¥ Second book by Tufte. Also well worth reading. Both are a visual delight!
¥ Statistics without Tears, Derek Rowntree. Penguin Books.
¥ Going on third party recommendation here, will try to get hold of a copy soon.
¥ Statistics for the Terrified, Gerald Kranzler and Janet Moursund.
Prentice Hall, Englewood Cliffs, NJ. 1995. (164 pages)
¥ Interesting to note how this book addresses its aims: (i)Êshort chapters (ii) Êlots of
examples (iii)Êpersonal language (iv)Êdoesn't attempt to give reasons for formulae.
¥ÊIÊliked chapter 7 'introduction to inferential statistics' which gives concepts first.
¥ÊThere are one or two minor problems (see 'who's afraid of ...' section above). ¥ÊAlso
interesting to note that the appendix on 'overcoming math anxiety' is mainly about
general techniques to deal with debilitating anxiety, but doesn't take the chance to make
the solutions specific for mathematics. It seems we all have trouble when too close to our
own areas of expertise.

Tables:
¥ Statistical Tables, F.D.J. Dunstan, A.B.J. Nix and J.F. Reynolds. R.N.D
Publications, Cardiff. ISBN 0-9506719-0-8. 1979. (67 pages)
¥ Elementary Statistical Tables, Henry R. Neave. Allen & Unwin, London.

Avoiding Damned Lies Ð Understanding Statistical Ideas Alan Dix www.meandeviation.com 26

 General (as opposed to 'statistics for ...') statistics textbook:
¥ Elementary Statistics 2e, Neil Weiss. Addison Wesley, Reading, MA.
1993. (733 pages)
¥ Good double page chapter start spreads including a mini-biography (I love them!) and
the introduction of a case study. The case study is revisited at the end of the chapter and
'solved' using the techniques introduced givings a sense of purpose. ¥ Chapter ends also
good with review of key terms and formulae and exercises, including some using
MINITAB. ¥ The book has a bit of a programmed learning feel which I've never got on
with personally. ¥ Great reference card (even better if it referenced text pages). ¥ Useful,
(but incomplete) procedure index on inside back cover. ¥ Example data sets listed in an
appendix and used in exercises.

Now the 'statistics for ...' books from Prentice Hall

¥ The Essence of Statistics for Business 2e, Michael C. Fleming and Joseph
G. Nellis. Prentice Hall, Hemel Hempstead 1996. (270 pages)
¥ Won hands down in my 'look up regression' test. ¥ 'Use of MINITAB' section at end of
each chapter. ¥ Short punchy (perhaps not so easy?) treatment with plenty of diagrams
and examples. ¥ 'Key learning point' section at end of each chapter summarising main
definitions and equations.
¥ Statistics for the Social Sciences Victoria L. Mantzopoulos. Prentice Hall,
Englewood Cliffs, NJ. 1995. (382 pages)
¥ Second in my 'look up regression' test. ¥ Separate study guide (I haven't seen it)
includes use of SPSS and MYSTAT. ¥ Grey boxes for key formulae and definitions.
¥ÊList of hypothesis testing formulae in appendix (not regression!). ¥ Dictionary of terms
and formulae in appendix (but no reference back into the text!). ¥ Ignore the shape of the
Chi squared curves in chapter 12 Ð the graphic artist obviously didn't know any statistics!
¥ Basic Statistics for the Social and Behavioural Sciences George M.
Diekhoff. Prentice Hall, Englewood Cliffs, NJ. 1996. (448 pages)
¥ Despite my tale of woe looking up regression formula, it is quite a nicely laid out book
with good use of two colour and lots of good features. ¥ÊGreat introduction giving an
overview of all the key statistical ideas in the book. ¥ Pink boxes for key formulae and
definitions. ¥ÊKeypoints highlighted in sidebars. ¥ÊFront and back covers used to list key
formulae (no page number reference back into the text, but equation numbers given.) ¥
Separate SPSS workbook. ¥ÊAppendix dedicated to explaining use of summation symbol.
¥ÊWorked examples at end of book. ¥ÊGlossary Ð but no page references!
¥ A First Course in Business Statistics 6e, James T. McClave and P. George
Benson. Prentice Hall, Englewood Cliffs, NJ. 1995. (746 pages)
¥ Full data set available separately on floppy. ¥ Generic (i.e. program independent)
'Using the Computer' exercises at the end of each chapter (also section 10.8 'Using the
Computer for Regression' telling you how to do it in SAS). ¥ Many SAS and some SPSS
printouts used in text. ¥ Floppy disk enclosed with book with ASP the student version of
a statistical package ... sounds good? Unfortunately, although there is a getting started
ASP tutorial in an appendix there is no other mention in the book. ¥ (Very) brief mention
of Baysian statistics.

Avoiding Damned Lies Ð Understanding Statistical Ideas Alan Dix www.meandeviation.com 27

 Books for scientists and engineers. Note that engineers and 'hard' scientists are
assumed to have more initial mathematics.
¥ Statistics for the Biosciences, William P. Gardiner. Prentice Hall, Hemel
Hempstead 1997. (416 pages)
¥ Subtitled 'data analysis using minitab software', the book does use MINITAB
extensively throughout. ¥ Appendix of 'statistical formulae' ... you guessed it, no
reference back to page numbers in the text!. ¥ Every chapter ends '...Êin the biosciences',
or '...Êfor biological experimentation', so you can't forget what book you're reading. ¥ Lots
on experimental design.
¥ Introductory Statistics for Environmentalists, Paul Moore and John
Cobby. Prentice Hall, Hemel Hempstead. 1998. (250 pages)
¥ Useful chapter on 'survey methods', which is not (as one might imagine) about
transepts of mud flats, but actually about ideas of different kinds of variable, random
sampling etc. ¥ Back cover says 'One of the few texts to make reference to MINITAB for
Windows and Excel for Windows'. Although it makes more mention of Excel than any
other book I looked at (not difficult), you'd better not blink.
¥ Statistics for Engineering and the Sciences 4e, William Mendenhall and
Terry Sincich. Prentice Hall, Englewood Cliffs, NJ. 1995. (1182 pages)
¥ Second book with ASP floppy (did Prentice Hall get a job lot cheap?) ... this time no
mention whatsoever of ASP software in text with exception of 'supplements' list in
preface which says the floppy is included. ¥ÊComputer lab section at the end of each
chapter giving work-through in SAS and MINITAB. ¥ÊLarge data sets listed in appendix
and available separately on floppy (why not include this rather than ASP!). ¥ Brief
mention of Baysian statistics.
¥ Miller and Freund's Probability and Statistics for Engineers 5e, Richard
A. Johnson. Prentice Hall, Englewood Cliffs, NJ. 1994. (630 pages)
¥ Some MINITAB exercises. ¥ Chapter on reliability and testing ¥ Special tables for
engineering data (e.g. sample size code letters from MIL-STD-105D) ¥ Short example of
Baysian statistics.
¥ Statistics for Analytical Chemistry 3e, J. C. Miller and J. N. Miller. Ellis
Horwood, Chichester. 1994. (233 pages)
¥ Assumes good basic knowledge of probability and maths. ¥ Treatment of issues such
as outliers in data and propagation of experimental errors during subsequent
computation. ¥ Lots of specialised issues: control charts, use of regression for calibration.
¥ Last chapter includes use of optimisation techniques for finding optimal factors and
pattern recognition techniques.

Books I've used myself Ð heavy mathematics!

¥ Mathematical Statistics 4e, John E. Freund and Ronald E. Walpole
Prentice Hall, Englewood Cliffs, NJ. 1987. (511 pages)
¥ Mathematical derivations of many common statistics.
¥ Theoretical Statistics, D. R. Cox and D. V. Hinkley. Chapman and Hall,
London. 1974. (608 pages)
¥ More fundamental again, doesn't deal with specific tests and statistics, but general
principles such as least-squares estimates, Baysian methods etc..

Avoiding Damned Lies Ð Understanding Statistical Ideas Alan Dix www.meandeviation.com 28