Design of experiments

Design of Experiments (DOE) is a systematic method for planning,

conducting, analyzing, and interpreting controlled tests to evaluate the
factors that may influence a particular outcome or response.

This method allows researchers to understand the

relationships between multiple input variables and key
output variables, leading to more efficient and effective
experimentation compared to traditional methods.
Components of DOE
1.Planning: This involves defining the objectives of the
experiment, selecting the factors to be studied, and
determining the levels of these factors.
2.Conducting: The experiment is carried out under
controlled conditions to ensure that the results are valid
and reliable.
3.Analyzing: Statistical methods are applied to analyze
the data collected from the experiments. This helps in
identifying significant factors and interactions.
4.Interpreting: The final step involves drawing
conclusions from the analysis, which can inform decision-
making and further research.
Steps in Designing an Experiment
1.Define the Objective: Clearly state the purpose of the experiment
and the questions to be answered.
2.Select Factors and Levels: Identify which factors will be tested and at
what levels.
3.Choose a Design: Depending on the complexity, choose between
different designs like:
1. Full Factorial: All possible combinations of factors and levels.
2. Fractional Factorial: A subset of the full factorial design to reduce the
number of experiments.
3. Response Surface Methodology (RSM): For optimizing responses with
multiple factors.
4.Randomization: Plan how to randomly assign the treatments to avoid
bias.
5.Replication and Blocking: Decide on the number of replications
needed and whether to implement blocking.
6.Conduct the Experiment: Carry out the experiment according to the
design.
7.Analyze the Data: Use statistical methods to analyze the data
collected, often involving ANOVA (Analysis of Variance) to assess the
effects of factors.
8.Interpret Results: Draw conclusions from the analysis, determine the
significance of factors, and make recommendations.
9.Validate Findings: If necessary, conduct additional experiments to
validate the results.
DOE Methods
1.Full Factorial Design
2.Fractional Factorial Design
3.Response Surface Methodology (RSM)
4.Taguchi Methods
5.Central Composite Design (CCD)
6.Box-Behnken Design
7.Historical Design
Factorial Design
Factorial design is a robust statistical method used in experimental
research to study the effects of two or more independent variables
(factors) on a dependent variable.
This approach allows researchers to investigate not only the main
effects of each factor but also the interaction effects between factors.
1.Factors and Levels:
1. Factors: Independent variables in an experiment that are
manipulated to observe their effect on a response variable. Each
factor can have multiple settings or values, known as levels.
2. Levels: The specific values or settings that a factor can take. For
example, a temperature factor might have levels of 50°C, 100°C,
and 150°C.
2.Runs:
1. Runs: The individual experimental trials or treatment combinations.
Each run represents a unique combination of factor levels.
Types of Factorial Designs
1.Full Factorial Design: This involves testing all possible
combinations of factor levels. While comprehensive, it can
be resource-intensive, especially with a large number of
factors.
2-Level Full Factorial Design: Each factor has two levels,
resulting in 2k runs, where k is the number of factors.
For example, a 2×2 factorial design with two factors each at
two levels results in four runs.
3-Level Full Factorial Design: Each factor has three
levels, resulting in 3k runs. This design allows for the study
of quadratic effects and interactions. For example, a 3×3
factorial design with two factors each at three levels results
in nine runs.
Example 1: Drug Dosage and Administration Route
Objective: Evaluate the effects of different drug dosages and
administration routes on patient outcomes.
Level Factor 1 Factor 2
Factors:
1.Dosage Levels: Low Dose Oral Intravenous

1. Low Dose (A1) High dose Oral Intravenous

2. High Dose (A2)
2.Administration Route: Full factorial design (22 factorial design)
1. Oral (B1) Experiment Factor 1 Factor 2
number
2. Intravenous (B2)
1 Oral (-) Intravenous(-)
2 Oral(+) Intravenous(-)
3 Oral(-) Intravenous(+)
4 Oral(+) Intravenous(+)
Full factorial design(23)
In an experimental investigation, three variables A,B and C. Lower level
and upper level variables are coded as’-’ and ‘+’ respectively.

Experimen A B C AB AC BC ABC
t no
1 - - -
2 + - -
3 - + -
4 + + -
5 - - +
6 + - +
7 - + +
8 + + +
Geometric view of two-level factorial designs: (a) two factors
and (b) three factors.
Fractional Factorial Design
It is a powerful experimental design technique used in statistics
and research to study the effects of multiple factors on a response
variable while minimizing the number of experimental runs.
This approach is particularly useful when dealing with a large
number of factors, as it allows researchers to focus on the most
significant effects without the need for a full factorial design, which
can become impractical due to the exponential increase in the
number of runs required.
Fractional factorial designs are often denoted as 2k-p where k is the
number of factors and p indicates the fraction of the full factorial
design used.
For example, when examining four input factors, a half factorial
design 24-1 = 8experiments may be implemented.
A quarter fraction factorial design 25-2 = 8 experiments may be
adopted to study 5 input factors.
Loss of resolution – Aliase
Resolution – a measure of confounding
 2III3-1 Design matrix 2IV 4-1 Design matrix
X1 X2 X3 X1 X2 X3 X4
1 -1 -1 +1 1 -1 -1 -1 -1
2 -1 +1 -1 2 +1 -1 -1 +1
3 +1 -1 -1 3 -1 +1 -1 +1
4 +1 +1 +1 4 +1 +1 -1 -1
5 -1 -1 +1 +1
6 +1 -1 +1 -1
7 -1 +1 +1 -1
8 +1 +1 +1 +1

Where,
X3=X1*X2 (X3 is confounding with X1 and X2)
X4=X1*X2*X3
Mixed Factorial Design:
This design involves at least one independent variable manipulated
within subjects and at least one between subjects.
It combines the strengths of both within-subjects and between-
subjects designs.
1.Between-Subjects Factors: These are factors where different
groups of participants are exposed to different levels of the factor.
For example, if you are studying the effect of a new teaching
method, one group of students might receive the new method
(treatment group), while another group continues with the traditional
method (control group).
2.Within-Subjects Factors: These are factors where the same
participants are exposed to all levels of the factor. For instance, if
you are measuring the effect of different types of study
environments (quiet vs. noisy), each participant would experience
both environments.
 Advantages of Full factorial design
1. Efficiency
Fewer Experiments: By examining multiple factors at once, researchers
can gather more information from fewer experimental runs compared to
studying one factor at a time.
2. Interaction Effects
Understanding Interactions: Factorial design allows researchers to assess
not only the main effects of each factor but also how they interact with
each other, providing a deeper understanding of the system being studied.
3. Comprehensive Analysis
Detailed Insights: It helps in identifying not just individual effects but also
combinations that might lead to unexpected results, which can be crucial
for applications like pharmaceuticals.
4. Randomization
• Reduced Bias: Randomization of treatment assignments helps control
for potential confounding variables, leading to more reliable results.
5. Flexibility
• Adaptable to Different Scenarios: Factorial designs can be easily
adapted to various fields, including agriculture, pharmaceuticals, and
manufacturing, making them widely applicable.
6. Statistical Power
• Increased Power: More comprehensive data collection can increase
the statistical power of the analysis, making it easier to detect
significant effects.
7. Robustness
• Exploring Robustness: Factorial designs help in assessing how robust
findings are to changes in experimental conditions, which is critical in
fields like drug development.
8. Cost-Effective
• Resource Efficiency: By optimizing the number of experiments,
factorial designs can reduce costs associated with time, materials, and
labor.
9. Facilitates Screening
• Screening Multiple Factors: Researchers can screen multiple factors
and levels in early phases of research to identify the most promising
ones for further study.
 Surface response methodology(RSM)
Response Surface Methodology is a statistical and
mathematical technique used for modeling and optimizing
complex relationships between multiple independent variables
(factors) and one or more dependent variables (responses). It is
particularly useful in situations where the relationship between
the factors and the response is not well understood or is nonlinear.
Advantages of RSM
• Efficient Exploration: Allows for the investigation of multiple factors
simultaneously.
• Predictive Modeling: Helps create predictive models for complex systems.
• Optimization: Facilitates finding optimal conditions for desired outcomes.
• Visualization: Provides graphical representations of the relationships between
factors and responses.
For example, the growth of a plant is affected by a certain amount of
water x1 and sunshine x2. The plant can grow under any combination
of treatment x1 and x2. Therefore, water and sunshine can vary
continuously. When treatments are from a continuous range of values,
then a Response Surface Methodology is useful for developing,
improving, and optimizing the response variable. In this case, the plant
growth y is the response variable, and it is a function of water and
sunshine. It can be expressed as
y = f (x1, x2) + ε
The variables x1 and x2 are independent variables where the response
y depends on them.
The dependent variable y is a function of x1, x2, and
the experimental error term, denoted as ε.
The error term ε represents any measurement error on the response,
as well as other type of variations not counted in f.
It is a statistical error that is assumed to distribute normally with zero
mean and variance σ 2 .
In most RMS problems, the true response function f is unknown. In
order to develop a proper approximation for f, the experimenter
usually starts with a low-order polynomial in some small region.
If the response can be defined by a linear function of independent
variables, then the approximating function is a first-order model.
A first-order model with 2 independent variables can be expressed as
Y= β0 + β1X1 + β2X2 + ε
If there is a curvature in the response surface, then a higher degree
polynomial should be used.
The approximating function with 2 variables is called a second-order
model:
Y= β0 + β1X1 + β2X2 + β2X212+ β11X212 + β22X222 + β12X1X2 + ε
In general all RSM problems use either one or the mixture of the both
of these models. In each model, the levels of each factor are
independent of the levels of other factors.
When the levels of each factor are not independent then a mixture
model is appropriate for designing an RMS model.
The response surface analysis is performed by using the fitted surface.
The response surface designs are types of designs for fitting response
surface.
Therefore, the objective of studying RSM can be accomplish by
(1) Understanding the topography of the response surface (local
maximum, local minimum, ridge lines), and
(2) Finding the region where the optimal response occurs. The goal is
to move rapidly and efficiently along a path to get to a maximum or
a minimum response so that the response is optimized.
 Response Surface Methods and Designs
Response Surface Methods are designs and models for working with
continuous treatments when finding the optima or describing the
response.
The first goal for Response Surface Method is to find the optimum
response. When there is more than one response then it is important
to find the compromise optimum.
The second goal is to understand how the response changes in a given
direction by adjusting the design variables.
The graph is helpful to see the shape of a response surface; hills,
valleys, and ridge lines. Hence, the function f (x1, x2) can be plotted
versus the levels of x1 and x2.
 y = f (x1, x2) + ε
In this graph, each value of X1 and X2 generates a y-value.
This three-dimensional graph shows the response surface from the side and it is called a
response surface plot.
The contour plots can show contour lines of X1 and X2 pairs that have the same response
value y.
When there are more than two independent variables, graphs are difficult or almost
impossible to use to illustrate the response surface, since it is beyond 3-dimension.
The most common designs used in response surface methodology are
Central composite design(CCD),
Box-Behnken design,
Dohlert design and
Mixture response surface methodology.
 Central composite Design
Central Composite Design (CCD) is a type of experimental design used
in Response Surface Methodology (RSM) to build a second-order
(quadratic) model for the response variable without needing a full
three-level factorial experiment.
CCD is particularly effective for optimizing processes and exploring the
relationships between factors.
This design has three different design points namely,
Edge point (±1)
Star point (±α)
Centre point
The center of the design space to a factorial point is ±1 unit for each
factor.
The distance from the center of the design space to a star point is |α| >
1.
The distance of each axial point (star point) from the centre in a CCD is α.
A value <1 puts the axial point in the cube,
A value = 1 puts them on the faces of the cube and
A value >1 puts them outside the cube.
The precise value of α depends on certain properties desired for the
design and on the number of factors involved.
• Types of central composite design
• The Box and Wilson design or CCD model comprising of factorial, and
factorial3 design.
• The star point outside the domine and the center point, representing the
experimental domine, helps determine the response surface plot.
• By estimating the precision of surface responses, the value of α can be
determined; where star design is ± α.
• There are three types of CCD; the α can be determined according to the
calculation possibilities and the required precision, which can be obtained
from surface responses.
• The α value’s positioning determines the quality of the design or
estimation. The rate by design is identified by determining the position of
the points.
• Eventually, the center composite design was classified into three types:
Central Composite
Design Type Terminology Comments

CCC designs are the original form of the central composite design. The star
points are at some distance from the center based on the properties
desired for the design and the number of factors in the design. The star
points establish new extremes for the low and high settings for all factors.
circumscribed CCC (5 levels)
These designs have circular, spherical, or hyperspherical symmetry and
require 5 levels for each factor. Augmenting an existing factorial or
resolution V fractional factorial design with star points can produce this
design.

For those situations in which the limits specified for factor settings are
truly limits, the CCI design uses the factor settings as the star points and
CCI (5 levels) creates a factorial or fractional factorial design within those limits (in other
Inscribed
words, a CCI design is a scaled down CCC design with each factor level of
the CCC design divided by α to generate the CCI design). This design also
requires 5 levels of each factor.

In this design the star points are at the center of each face of the factorial
CCF (3 levels) space, so α = ± 1. This variety requires 3 levels of each factor. Augmenting
Face Centered
an existing factorial or resolution V design with appropriate star points can
also produce this design.
 Experimental run using CCD

Experiment Component Factor X1 Factor X2

run
1 Centre point 0 0
2 Factorial point -1 -1
3 Factorial point -1 +1
4 Factorial point +1 +1
5 Factorial point +1 -1
6 Star point -α 0
7 Star point +α 0
8 Star point 0 +α
9 Star point 0 -α
Eg: The experimental design layout developed by CCD
Formulations Oil(ml) Smix(ml) Globule size PDI Transmittance(%)
F1 0 0 98 0.82 97.58
F2 0 0 85 0.85 96.34
F3 0 0 102 0.78 9708
F4 1.41421 0 11 0.46 99.57
F5 -1 -1 81 0.64 96.95
F6 1 1 129 0.76 99.58
F7 0 0 89 0.98 96.97
F8 1 -1 165 0.64 97.47
F9 -1 1 59 0.57 98.86
F10 0 0 95 0.89 94.5
F11 0 -1.41421 135 0.81 95.42
F12 -1.41421 0 58 0.51 98.54
F13 0 1.41421 106 0.67 97.47
Box-Behnken design
Box-Behnken design is an experimental design used in statistics which is
devised by George E.P. Box and Donald Behnken in 1960.
It is a second order quadratic model for non-linear responses.
It is mainly used when factor space is within the experimental domain
which differentiate BBD and CCD.
In this design the treatment combinations are at the midpoints of edges
of the process space and the centre.
The designs are rotatable and it uses the twelve middle edge nodes and
three centre nodes to fit a second order equation.
The number of experiments(N)
N= 2K(K-1)+C0
Where, K= number of factors, C0 = Number of central points
Three variable Box-Behnken design
Run X1 X2 X3
1 -1 -1 0
2 -1 1 0
3 1 -1 0
4 1 1 0
5 -1 0 -1
6 -1 0 1
7 1 0 -1
8 1 0 1
9 0 -1 -1
10 0 -1 1
11 0 1 -1
12 0 1 1
13 0 0 0
14 0 0 0
15 0 0 0
Historical Design

Historical research is a qualitative technique. Historical research

studies the meaning of past events in an attempt to interpret the facts
and explain the cause of events, and their effect in the present
events.
The aim of the historical research is to show the importance of past
event in present situation.
While writing a historical research report , the actual events and the
conditions of the time are not violated, exaggerated or distorted.
Characteristics
• Historical research involves careful study and analysis of data about
past events.
• Historical research is a critical investigation of events , their
development ,experiences of the past.
• Historical research involves the review of written materials ,but may
include oral documentation as well.
• Historical research typically relies on available data , Data for historical
research are usually in the form of written narrative records of the
past , diaries ,letters ,newspapers , minutes of meetings , reports, and
so on.
• Historical research relies on a wide variety of sources, both primary &
secondary including unpublished material.
Primary Sources
 Eyewitness accounts of events
 Can be oral or written testimony
 Found in public records & legal documents, minutes of meetings,
corporate records, recordings, letters, diaries, journals, drawings.
 Located in university archives, libraries or privately run collections
such as local historical society.
Secondary Sources
 Can be oral or written
 Second hand accounts of events
 Found in textbooks, encyclopaedias, journal articles, newspapers,
biographies and other media such as films or tape recordings.
Steps of historical design
Identify an idea, topic or research question
Conduct a background literature review
Refine the research idea and questions
Determine that historical methods will be the method used
Identify and locate primary and secondary data sources Evaluate the
authenticity and accuracy of source materials Analyse the date and
develop a narrative exposition of the findings.
Areas of historical study
Periods:
Historical studies often focus on events and developments that occurred
during particular blocks of time in the past . Historical researchers gives
these periods of time names in order to allow the organization of ideas and
the classificatory generalization to be used by these historical researchers.
Geographical locations:
Particular geographical locations can form the basis of historical study eg.
countries and cities.
Military history :
Military history concentrates on the study of conflicts that have
happened in human society. This includes examining the wars, battles,
military strategies and weaponry.
Optimization Techniques
1.Gradient Descent:
1. An iterative optimization algorithm used to minimize the response surface.
2. Starts with an initial guess and iteratively adjusts to find the minimum.
2.Genetic Algorithms:
1. A search heuristic inspired by natural selection.
2. Uses techniques like selection, crossover, and mutation to explore the
solution space.
3. Suitable for complex optimization problems with multiple objectives.
3.Simulated Annealing:
1. A probabilistic technique for approximating the global minimum of a function.
2. Inspired by the annealing process in metallurgy.
3. Balances exploration and exploitation to escape local minima.
4. Response Optimization:
1. Involves adjusting factors to achieve optimal response levels.
2. Often employs methods like desirability functions to convert multiple
responses into a single score.
5. Multi-Objective Optimization:
3. Deals with problems involving more than one objective function.
4. Techniques like Pareto optimization help identify trade-offs between
conflicting objectives.