Experimental Design and Analysis

2.1 Experimental Design

 Design of experiment(DOE) has its own terminology,

methodology and subject of research.

 DOE is a planned approach for determining cause and effect

relationships

 It clearly indicates that it deals with experimental methods.

 A large number of experiments is done in research,

development and optimization of the system.

12/30/2021 10:21:47 AM 1
 Cont’d
 The model usually differs from the subject in its dimensions and
sometimes in its nature.

 An experiment takes a central place in science, particularly

nowadays, due to the complexity of problems science deals with.

 To increase research efficiency, it is necessary to introduce

something completely new into classical experimental research.

 One kind of innovation could be, to apply statistical mathematical

methods or to develop design of experiments (DOE).

 .

12/30/2021 10:21:47 AM 2
 Essentials of Experimental Design

 Reduction or minimization of total number of trials,

 Simultaneous varying of all factors that formalizes

experimenter’s activities, and

 Choice of a clear strategy that enables reliable

solutions to be obtained after each sequence of
experiments.

12/30/2021 10:21:47 AM 3
Experimental work is included in:

 physical and chemical studies for establishing

constants and properties of elements, chemical

compounds and materials,

 routine analyses of raw materials,

intermediates and final products,

12/30/2021 10:21:47 AM 4
 lab studies for designing and developing
technological processes,

 optimization of technological procedures in the

lab, pilot-plant and full-scale plant systems,

 optimization of mixture,

 mathematical modeling of a system, and

 selection of factors by the significance of their

effects on a measure value response
12/30/2021 10:21:47 AM 5
 The Importance of Careful Experimental Design
 Experimental design is a careful balancing of several features
including power, validity, practicality and cost.
 Our goal is always to actively design an experiment that has the
best chance to produce meaningful, defensible evidence, rather
than hoping that good statistical analysis may be able to correct
for defects after the fact.
Power – meaningful result
Validity- defensible evidence, reliable result
Practicality- measurable and attainable
Cost- cost effective & affordable

12/30/2021 10:21:47 AM 6
 Types of experimental design
There are different types of experimental design
approaches. Here we will discuss only the
following two types:

i. Full factorial experimental design

ii. Fractional factorial experimental design

iii. RSM –CCD

12/30/2021 10:21:47 AM 7
 Full factorial experimental design

For full factorial experimental design, the total number of

expected experiments is given by:
N=kyn
Where:
N is total number of experiments.
k is number of replica.
y is number of levels.
n is number of factors.
12/30/2021 10:21:47 AM 8
 con’t
Example 1: A student wants to produce ethanol from sugar
cane bagasse. Among all factors which will affect the
fermentation process, his/ her aim is to determine the effect
of temperature(oC), molasses concentration(Bo) and size of
inoculums(%) of yeast. For full factorial experimental
design:
a) How many factors will be determined by the students?
b) For two levels and three replica, how many number of
experiments will be expected?
c) For three levels and three replica, how many number of
experiments will be expected?

12/30/2021 10:21:47 AM 9
 Cont’d
Advantages & disadvantages of full factorial
experimental design
Advantages:
 It minimizes errors since it accounts the effects of all
factors
 It provides reliable data to determine the optimum
values and to compare the effects of each factor and
their interaction.
Disadvantages:
 It increases working time
 It makes difficult to manage the experiments
 It increases cost of experiments.
12/30/2021 10:21:47 AM 10
 Fractional Factorial Experimental Design
For fractional factorial experimental design, the total number of
expected experiments is given by:

N=kyn-1

Where:

N is total number of experiments.

k is number of replica.

y is number of levels.

n is number of factors.

12/30/2021 10:21:47 AM 11
 Cont’d
Example 2: A student wants to produce ethanol from sugar cane
bagasse. His/ her aim is to determine the effects of
temperature(oC), molasses concentration(Bo) and size of
inoculums(%) of yeast on fermentation process. For fractional
factorial experimental design:
a) How many factors will be determined by the students?
b) For two levels and three replica, how many number of
experiments will be expected?
c) For three levels and three replica, how many number of
experiments will be expected?
d) For fractional factorial experimental design, what can he/she
ignore one factor among others to determine the effect of two
factors ?

12/30/2021 10:21:47 AM 12
 Con’t
Advantages & disadvantages of fractional factorial
experimental design
Advantages:
 It saves working time
 It makes easy to manage the experiments.
 Reduces costs of experiments
Disadvantages:
 It brings errors since it doesn’t account the effects of
all factors
 It doesn’t provide reliable data to determine the
optimum values and to compare the effects of each
factor and their interaction.
12/30/2021 10:21:47 AM 13
 Arrangement of Experimental Design
 One you have designed your experimental works, you
have to arrange your experimental activities with
measureable qualities of the factor levels & replica.

 When you assign the levels, their interval shall be

consistent and selected with scientific evidences, based on
pervious research works.

 If you assign the levels without scientific justification,

your data will not be reliable.

12/30/2021 10:21:47 AM 14
 Con’t
Example 3: A student wants to produce ethanol from sugar cane
bagasse. Among all factors which will affect the fermentation
process, his/ her aim is to determine the effect of temperature from
25-35 oC), molasses concentration from 25- 35 Bo and size of
inoculums from 20 - 40 %v/v of yeast. For full factorial
experimental design:
a) For two levels and two replica, show the arrangement of number
of experiments will be expected.
b) For two levels and three replica, show the arrangement of number
of experiments will be expected.
c) For three levels and two replica, show the arrangement of number
of experiments will be expected.
d) For three levels and three replica, show the arrangement of
number of experiments will be expected.

12/30/2021 10:21:47 AM 15
 Cont’d
Solutions:
a) For two levels & two replica, N=2*2^2= 16
Temperature= 25-35oC,; T1= 25 oC, T2= 35oC
Molasses concentration= 25-35Bo; C1=25Bo, C2= 35Bo
Size of inoculums= 20-40%v/v; S1= 20%v/v, S2= 40%v/v

12/30/2021 10:21:47 AM 16
 Cont’d
Solution:
b) For two levels & three replica, N= 3*2^2=24
Temperature= 25-35oC,; T1= 25 oC, T2= 35oC
Molasses concentration= 25-35Bo; C1=25Bo, C2= 35Bo
Size of inoculums= 20-40%v/v; S1= 20%v/v, S2= 40%v/v

12/30/2021 10:21:47 AM 17
 Cont’d
Solutions:
c) For three levels & two replica, N= 2*3^3=54
Temperature= 25-35oC,; T1= 25 oC, T2= 30oC, T3= 35oC
Molasses concentration= 25-35Bo; C1=25Bo, C2= 30Bo, C3= 35Bo
Size of inoculums= 20-40%v/v; S1= 20%v/v, S2= 30%v/v, S3= 40%v/v

12/30/2021 10:21:47 AM 18
 Cont’d

12/30/2021 10:21:47 AM 19
 Cont’d

12/30/2021 10:21:47 AM 20
 Cont’d
Solution:
c) For three levels & two replica, N= 3*3^3=81
Temperature= 25-35oC,; T1= 25 oC, T2= 30oC, T3= 35oC
Molasses concentration= 25-35Bo; C1=25Bo, C2= 30Bo, C3= 35Bo
Size of inoculums= 20-40%v/v; S1= 20%v/v, S2= 30%v/v, S3= 40%v/v

12/30/2021 10:21:47 AM 21
 Cont’d

12/30/2021 10:21:47 AM 22
 Cont’d

12/30/2021 10:21:47 AM 23
 Cont’d

12/30/2021 10:21:47 AM 24
 cont’d

Example 4: A student wants to produce biodiesel from non-

edible oil. His/ her aim is to determine the effect of
temperature from 55-75 oC, alcohol to oil ratio 6:1 – 10:1 and
amount of NaOH as catalyst is 0.5-1.0 %w/w on
transesterification reaction . For full factorial experimental
design:
a) For two levels and three replica, show the arrangement of
number of experiments will be expected.
b) For three levels and three replica, show the arrangement of
number of experiments will be expected.
NB: Homework exercise

12/30/2021 10:21:47 AM 25
 Cont’d
Example 5: A student wants to produce
biodiesel from non-edible oil. His/ her aim is
to determine the effect of temperature from 55-
75 oC, alcohol to oil ratio 6:1 – 10:1 and
amount of NaOH as catalyst is 0.5-1.0 %w/w
on transesterification reaction . For response
surface-central composite experimental
design, show the arrangement of number of
experiments will be expected.

30/12/2021 26
 Solution

30/12/2021 27
 Effect determination
For full factorial experimental design, the effect of each factor
and their interaction is determined by:

Where: X is the factor, A,B,C……

R is the response / product/ output
y is number of levels
x+ is the maximum value of the factor, A, B, C…
x- is the minimum value of the factor, A, B, C…

12/30/2021 10:21:47 AM 28
 2.2. Experimental Data Analysis
Exploratory Data Analysis(EDA)
 Statistical analysis of experiments starts with graphical
and non-graphical exploratory data analysis (EDA).
EDA is useful for:
 Detection of mistakes;
 Checking of assumptions;
 Determining relationships among the explanatory
variables;
 Assessing the direction and rough size of relationships
between explanatory and outcome variables; and
 Preliminary selection of appropriate models of the
relationship between an outcome variable and one or
more explanatory variables.
12/30/2021 10:21:47 AM 29
 Statics & Probability : Review
Measures of central tendency and dispersion
 The statistical term used for ‘average’ is the arithmetic mean.

 Mean, median and the modal values are measures of central

tendency

For discrete data:

Mean: The arithmetic mean value is found by adding together
the values of the members of a set and dividing by the number
of members in the set.

12/30/2021 10:21:47 AM 30
 cont’d
Median: The median value often gives a better indication of
the general size of a set containing extreme values.

The median value is obtained by:

i. ranking the set in ascending order of magnitude, and

ii. selecting the value of the middle member for sets

containing an odd number of members, or finding the
value of the mean of the two middle members for sets
containing an even number of members.

12/30/2021 10:21:47 AM 31
 cont’d

Mode: The modal value, or mode, is the most

commonly occurring value in a set. If two values
occur with the same frequency, the set is bi-modal.

Example 5: Determine the mean, median and mode for

the set:

{2, 3, 7, 5, 5, 13, 1, 7, 4, 8, 3, 4, 3}

12/30/2021 10:21:47 AM 32
 cont’d
Mean, median and mode for grouped data
Mean: The mean value for a set of grouped data is found by
determining the sum of the (frequency × class mid-point
values) and dividing by the sum of the frequencies,

where f is the frequency of the class having a midpoint value of x.

12/30/2021 10:21:47 AM 33
 Measures of dispersion
 Standard deviation and variance are measures of dispersion.
For discrete data:
Standard deviation(σ): The standard deviation of a set of data
gives an indication of the amount of dispersion, or the scatter,
of members of the set from the measure of central tendency.

 Its value is the root-mean-square value of the members of the

set and for discrete data is obtained as follows:
i. Determine the mean value
ii. Calculate the deviation of each member of the set from the
mean, giving

12/30/2021 10:21:47 AM 34
 cont’d
iii. determine the squares of these deviations,

iv. find the sum of the squares of the deviations, that is

v. determine the standard deviation,

12/30/2021 10:21:47 AM 35
 cont’d

For grouped data:

where f is the class frequency value, x is the

class mid-point value and x is the mean value
of the grouped data.
12/30/2021 10:21:47 AM 36
 Probability
The probability of something happening is the likelihood or chance of it
happening.
Laws of probability
The addition law of probability:
 The addition law of probability is recognized by the word ‘or’ joining the
probabilities.
 If pA is the probability of event A happening and pB is the probability of event
B happening, the probability of event A or event B happening is given by
pA+pB (provided events A and B are mutually exclusive, i.e. A and B are events
which cannot occur together).
 Similarly, the probability of events A or B or C or . . .N happening is given by:
pA+pB+pC+….+pN

12/30/2021 10:21:47 AM 37
 cont’d
The multiplication law of probability
 The multiplication law of probability is recognized by
the word ‘and’ joining the probabilities.
 If pA is the probability of event A happening and pB is
the probability of event B happening, the probability
of event A and event B happening is given by pA×pB.
Similarly, the probability of events A and B and C
and . . .N happening is given by:
pA ×pB ×pC × ・ ・ ・ ×pN
12/30/2021 10:21:47 AM 38
 Linear regression

 Regression analysis is the art and science of

fitting straight lines to patterns of data.

 In a linear regression model, the variable of

interest to relate dependent variables with
independent variable using a linear equation.

12/30/2021 10:21:47 AM 39
 Cont’d
For linear function of y= ao +a1x

12/30/2021 10:21:47 AM 40
 Worksheet
1. Explain briefly the difference between applied
researches and basic researches
2. Explain briefly the purposes of research
3. Explain briefly the importance of careful
experimental design
4. What are the motive factors to do researches?
5. Explain briefly the meaning of experimental
design.
6. List the structures of research proposal and
explain the purposes of each section.
7. List the structures of research report and explain
the purposes of each section.
30/12/2021 41
 Purposes of Data Analysis
The purpose of analyzing data is to obtain
usable and useful information.
The analysis, irrespective of whether the data
is qualitative or quantitative, may:
describe and summaries the data
identify relationships between variables
compare variables
identify the difference between variables
forecast outcomes
 Data interpretation
 Numbers don’t speak themselves
 Interpretation is the process of attaching meaning
to the data.
 Interpretation demands fair and careful
judgments. Often the same data may be
interpreted in different ways.
 So it is helpful to involve others or take time to
hear how different people interpret the same
information.
 Cont’d
 Before you interpret your data review basics of
principles of statistics and data interpretation tools.
 Review the scientific meanings of the values to
give justification.
 During statistical data analysis ask consultation
from statisticians for accurate interpretation.
 Use both analytical and software tools to analyze
your data depending on the nature of it.