BASIC PRINCIPLES OF EXPERIMENTAL DESIGNS

Professor R. A. Fisher stated the basic principles of experimental research during his pioneering work
at Rothamsted Experimental Station in the 1920s and 1930s.

These are:

(1) Principle of Replication

(2) Principle of Randomization

(3) Principle of Local Control

Principle of Replication states that the experiment should be repeated more than once. Each
treatment is applied in many experimental units instead of one; this increases the statistical accuracy
of the experiments. Thus, replication principle refers to the deliberate repetition of an experiment,
using an almost identical procedure with a different set of subjects, in a different setting and at a
different time. Winer (1971:391) has said, "A replication of an experiment is an independent
repetition under as nearly identical conditions as the nature of the experimental material will
permit." Replication is used to increase the precision of a study, to increase the accuracy with which
the main effects and interactions can be estimated.

Principle of Randomization helps to control the effect of extraneous factors by randomization.

Randomisation is used to make the test more valid by ensuring that the independence of the
observations are maintained when the samples are drawn randomly from the population or subjects
are randomly assigned to experimental treatments. The principle of randomization is applied to
estimate the experimental error better.

Principle of local control refers to the amount of balancing blocking and grouping of the subjects or
experimental units used in the experimental design. Through the principle of local control the
variability due to extraneous factors from the experimental error can be eliminated. In grouping
homogeneous subjects or experimental units are assigned two different groups so that the different
groups main receive differential experimental treatments for blocking refers to the assignment of
experimental unit two different block in such a way that is experimental unit within a block may be
homogeneous. Balancing in an experimental design may be explained as grouping blocking and
assignment of experimental units’ two different treatments so that the resulting design appears to
be a balanced one. The principle of local control makes their experimental design statistically and
experimentally accurate.

DESIGNS
Randomized
Two-Randomized-Groups Design

In a two-randomized-groups design the subjects are randomly assigned to two groups. The
independent variable and the dependent variables are first defined by the experimenter, and two
values of the independent variable are selected. These may be called 'conditions' or 'treatments’.
The main aim is to find out if the two conditions affect the dependent variable in a differential way.
In selecting a sample for the proposed experiment, the population is defined.
The experimental group is given one type of treatment and the control group is given another type
of treatment. Usually the test or its non-parametric counter-part, the Mann Whitney U test-is
applied in a two-randomized-groups design. Underwood (1966) stated two ways through which
unbiased groups or random groups of subjects can be formed: captive assignment and sequential
assignment.

In the technique of captive assignment all subjects are individually known to the experimenter by
name and they are all present at one time. In the techniques of sequential assignment subjects are
assigned into the various groups through complete randomization or block randomization.

ADVANTAGES:

The randomized two-group design is one of the simplest available, yet it has several advantages over
other, more complex designs.

It is economical in terms of time and materials as it involves few subjects.

No pretesting or categorization of subjects is necessary.

Finally, statistical analysis of the resulting data is relatively simple.

The Randomized-Multigroup Design

One way to expand the randomized twogroup design is to add one or more levels of the
independent variable.

A variation on the single-factor multigroup design is one that includes multiple control groups. Due
to some limitations of the two-randomized-groups design, the randomized twogroup design may be
used. There are three or more conditions or values of the independent variable and three or more
groups of subjects participate in the experiment. The subjects are randomly assigned to three or
more unbiased groups.

Matched group
In the matched-groups design all subjects are tested on a common task or a pretest measure (also
called the matching variable) first and then, they are formed into groups. This is done on the basis of
the performance of the common task or the matching variable. The groups are to be equivalent
groups. Then the different conditions or values of the independent variable are introduced to each
group. If these groups have equivalent means on the dependent variable, before the experimental
treatment is given and if a significant difference occurs after administering the experimental
treatment and controlling the relevant variables, then the differences may be attributed to the
experimental treatment.

This is done to ensure that all the groups have equal dependent variable values before the
administration of the experimental treatment.

The matching variable is one that yields a high correlation with the experimental task or the
dependent variable. If the matching variable is highly correlated with the scores on experimental
task or with scores on the dependent variable, the matching may be regarded as successful. Since
the matched subjects are out into different groups, the effect of the characteristic on which the
subjects were matched gets distributed evenly across the treatments. So, the characteristic
contributes little to the differences between group means

The method used for dividing each pair into two groups is randomization. This is done to avoid the
experimenter's bias in matching.

Advantages:

Subject variables that may have an effect on the independent variable under investigation can be
controlled, thus, matching can increase the experiment’s sensitivity to the effect of the independent
variable.

This design may also help to discover effects that would be otherwise overlooked.

A given effect can be demonstrated with fewer subjects, thus saving time and money.

The matched-pairs design is the matched-groups equivalent to the randomized two-group design.
Here only 2 groups are present, which makes the experiment less time consuming.

Matched-Multigroup Designs

In matched multigroup designs matched subject are required for every treatment group in the
experiment. If the experiment includes four treatment groups, quadruplets of subjects having similar
characteristics on the variables being matched would be required.

Ex-post facto
Ex post facto study or after-the-fact research is a category of research design in which the
investigation starts after the fact has occurred without interference from the researcher. In ex post-
facto research the effects of a naturally occurring treatment are evaluated after it has occurred. The
researcher does not control or manipulate the independent variable. Therefore, the treatment is
given by selection and not by manipulation. Due to this it is difficult to find out the cause-effect
relationship between the dependent variable and the independent variable.

The investigator attempts to trace an effect which has already occurred to its probable causes. The
effect becomes the dependent variable and the probable causes become the independent variable.
The outcome or the dependent variable measure is related with treatments that have already
occurred.

There are two common types of ex post facto design, namely, correlational design and criterion-
group design.

CHARACTERISTICS OF EX POST FACTO RESEARCH

Ex-post factor research is also known as ‘causal comparative research’. The ex-post facto research
has certain characteristic which distinguishes it from other different types of researches. Some of
these characteristics are presented below in.

The research has a control or a comparison group

As the research is done on basis of the study of the cause which has already led to its effects, it
becomes necessary for the researcher to keep a control group, which can be used for comparison
with the actual experimental group later on, in order to analyse the cause of an already occurred
event.

The behaviour, action, event or the treatment or the independent variable of the research cannot be
manipulated or changed

As the ex-post research is a kind of study which tries to predict the causes on the basis of actions
that have already occurred, the researcher cannot manipulate or change the already occurred
actions or behaviour.

The research focuses on the effects

Since the researcher tries to analyse and predict the reasons behind the occurrence of an event or
phenomena, their first attempt is to focus on the event or the phenomena that has already
occurred. Only after having a detailed study of the phenomena or the event, the researcher tries to
determine the causes behind such an event or phenomena.

The research tries to analyse the ‘how’ and ‘what’ aspect of an event Since the researcher tries to
understand the causal effects behind a phenomena, the research basically focuses on how and what
reasons that has led that phenomena to occur.

Explores possible effects and causes

With the help of an ex-post facto research, the researcher tries to analyse the cause and effect
phenomena of an event, action or behaviour.

DIFFERENCES BETWEEN AN EXPERIMENTAL EX-POST FACTO RESEARCH

Control over independent Variable

In an experimental research, the researcher can directly manipulate the independent variable/s (that
is, the cause) in order to examine its effect on the dependent variable (that is, the effect).

In an ex-post facto research, the researcher cannot directly manipulate the independent variable/s
(that is, the cause) as he or she predicts the cause on basis of the dependent variable (that is, the
effect).

Principle of randomisation

The researcher can use the principle of randomisation in the assignment of subject to different
groups or assignment of treatment to differently groups in an experimental research on basis of
which they can conclude or infer that other things remaining equal/ constant/controlled the effect is
a result of manipulation of the cause.

The researcher cannot use the principle of randomisation in an ex-post facto research as the
researcher has no direct control over the cause and so they infer the possibilities of the causes on
basis of the existing effect.

Interpretation

It is easier to interpret or infer relationships between the independent and dependent variables as
they can manipulate the independent variable and see its effect on dependent variable
It is difficult to interpret or infer relationship between the independent and dependent variables as
there can be more than one possibilities or cause for a particular effect.

Manipulation of variables

The researcher can manipulate variables in an experimental research.

The researcher cannot manipulate variables in an expost facto research

STEPS of EX POST FACTO

The sequence of steps involved in ex post facto research was set forth by Isaac and Michael (1971),
and they include:

1. Defining the problem: In an ex-post facto research, the researcher to should not only need to find
out a problem, but also need to determine, analyse and define the problem which they will be
dealing with.

2. Survey the literature: Before trying to predict the causal relationships, the researcher needs to
study all the related or similar literature and relevant studies, which may help in further analysis,
prediction and conclusion of the causal relationship between the variables under study.

3. State the hypotheses: The researcher should propose the possible solutions or alternatives that
might have led to the effect.

4. List the assumptions upon which the hypotheses and procedures will be based: the researcher
needs to list out the assumptions which will be the basis of the hypothesis and procedure of the
research

5. Design the approach:

a. Select appropriate subjects and source materials.

b. Select or construct techniques for collecting the data.

c. Establish categories for classifying data that are unambiguous, appropriate for the purpose of the
study, and capable of bringing out significant likenesses or relationships.

6. Validate the data-gathering techniques: The researcher needs to validate the significance of their
research. They need to be cautious regarding the extent to which their findings would be valid and
significant and helpful in interpreting and drawing inferences from the obtained results.

7. Describe, analyze, and interpret the findings in clear, precise terms: Finally, the researcher needs
to analyse, evaluate and interpret the information collected. On basis of this step, the researcher
selects the best possible alternative of causes which might have led the effect to occur

STRENGTHS AND WEAKNESSES OF EX-POST FACTO RESEARCH

No research can be perfect in itself. All methods have their strengths as well as weaknesses. The
same is applicable in the case of ex-post factor research too.

The strengths of the ex-post facto research are:

It is considered as a very relevant method in those behavioural researches where the variables
cannot be manipulated or altered. The examples of such researches can include many sociological
(e.g. delinquency) as well as educational variables ( e.g. achievements).

It is more useful than an experimental research as it can be used in analysing a cause on basis of the
effect, which is impossible in an experimental research.

It is less time consuming as well as economical.

Ex-Post Facto Research It gives a chance to the researcher to analyse on basis of his personal opinion
and then come out with the best possible conclusion.

The weaknesses as well as the limitations of the ex-post facto research are:

In an ex-post facto research, the researcher cannot manipulate the independent variables.

The researcher cannot randomly assign the subjects to different groups.

The researcher may not be able to provide a reasonable explanation for the relationship between
the independent and dependent variables under study

Correlational Design
A correlational approach (also known as the psychometric approach) is one in which the
experimenter collects two or more sets of data from the same group of subjects so that the
relationship between the two subsequent sets of data can be determined.

O1 O2

where O1 and O2 are 2

Criterion-group Design

In the criterion-group design, the experimenter tries to find out what caused the particular state of
condition by comparing the characteristics of the group which possesses the criterion behaviour
with those who do not.

C O1 O1 C O2 C O1

Where C is the criterion

However, this criterion-group study provides some testable hypotheses in terms of the potential
causes, which by using a quasi-experimental design or true experimental design may be tested in a
more scientific way.
Quasi Experimental design
According to Mcburney & White (2007) ‘quasi experiment is a research procedure in which the
scientist must select subjects for different conditions from preexisting groups’.

According to Broota (1989) “All such experimental situations in which the experimenter does not
have full control over the assignment of experimental units randomly to the treatment conditions or
the treatment cannot be manipulated are called quasi experimental design.”

According to Singh (1998) “A quasi experimental design is one that applies an experimental
interpretation to results that do not meet all the requirement of a true experiment.”

Characteristic of quasi experimental research design

Manipulation of independent variable

A quasi-experimental design aims to establish a cause-and-effect relationship between an

independent and dependent variable. Quasi experimental research design involves the manipulation
of independent variable to observe the effect on dependent variable. Quasi experimental designs
are generally used to establish the causality (effect of independent variable on dependent variable)
in situations where researchers are not able to randomly assign the subjects to groups for various
reasons.

Lack of one or two essential character of true experimental design

This design lacks the characteristics of control and randomization. In quasi experimental situation,
the researcher does not have control over the assignment of subject to condition. He cannot
randomly assign subjects to different groups or randomly assign treatments to different groups. The
Quasi experimental design can be characterized by a lack of control over extraneous variables.

Quasi independent variable used instead of true independent variable.

A Quasi experimental design uses an quasi independent variable rather than a true independent
variable. A quasi independent variable refers to a correlational variable that simulates an
independent variable in an experiment.

Validity

Quasi experimental research design has high external validity than most true experiments, because
they often involve real-world interventions instead of artificial laboratory settings. With respect to
internal validity, they often appear to be inferior to randomized experiments.

Difference between True and Quasi

Assignment to groups

In True experimental design the researcher randomly assigns subjects to control and treatment
groups.

In Quasi-experimental design some other, non-random method is used to assign subjects to groups.
 Control over treatment

The researcher usually designs the treatment and decides which subjects receive it in true
experimental design.

In quasi experimental design, the researcher often does not have control over the treatment, but
instead studies pre-existing groups that received different treatments after the fact.

Use of control groups

True experimental design requires the use of control and treatment groups.

In quasi experimental design control groups are not required (although they are commonly used).

Type of research

True experiments study a cause and effect relationship, where as quasi experiments are
correlational research design.

Advantages

 Quasi-experimental research may be more feasible because it often does not have the time and
logistical constraints associated with many true experimental designs.
 True experimental designs are sometimes impractical or impossible because the research can only
effectively be carried out in natural settings. Experimental research can create artificial situations
that do not always represent real-life situations. This is largely due to fact that all other variables are
tightly controlled which may not create a fully realistic situation. For this reason, external validity is
increased quasi-experimental research.
 Reactions of test subjects are more likely to be genuine because it not an artificial research
environment.
 It can be very useful in identifying general trends from the results, especially in social science
disciplines.
 It reduces the difficulty and ethical concerns that may surround the pre-selection and random
assignment of test subjects. For example, if examining the effects of cigarette smoking by pregnant
women on the fetus, it would be unethical to randomly assign pregnant women to groups.
 Matching procedures may be used to help create a reasonable control group, making generalization
more feasible. For example, if one group of migraine suffers received a new treatment and a second
group did not receive the treatment, the difference in the pain levels may be attributed to the
treatment if the control group is an appropriate comparison group.
 The results generated can often be used to reinforce the findings of case studies by conducting
research that may lend itself to statistical analysis.

Disadvantages:

 The lack of random assignment into test groups leads to non-equivalent test groups which can limit
the generalizability of the results to a larger population. Beside of the lack of randomization and the
reduced internal validity, conclusions about causality are less definitive in quasi-experimental
designs.
 Statistical analyses may not be meaningful due to the lack of randomization and the threats to
internal validity.
 Pre-existing factors and other influences are not taken into account because variables are less
controlled in quasi-experimental research. For example, when examining the impact of smoking by
 pregnant mothers, there may be other factors such as diet, education, overall health, and access to
health care in general that may be playing a role in the outcome. If other variables are not
controlled, the researcher can be assured that the treatment was the sole factor causing the
outcome.
 Human error also plays a key role in the validity of any project as discussed in previous modules.
 The research must adhere to ethical standards in order to be valid.

QUASI EXPERIMENT

A Quasi experimental design is one that resembles an experimental design but uses and quasi
independent variable rather than a true independent variable. A quasi independent variable refers
to a correlational variable that simulates an independent variable in an experiment. The
independent variable is not manipulated in this design. The Quasi experimental design can be
characterized by a lack of control over extraneous variables. Even though there is some amount of
control all extraneous variables cannot be control which may compromise the internal validity of the
experiment. In a Quasi experimental design control over when and whom the measurement is
applied can however since the subject and not randomly assigned to experimental and control group
therefore equivalence of the group is not maintained and this poses some threat to the validity of
the experiment.

Time series designs

When it is not possible to include a control or comparison group time series design may be used.
This helps to control the extraneous variables better. In the time series design (Campbell & Stanley,
1963) several observations of behaviour over time prior to the treatment and then immediately after
the treatment are made. Extraneous variables like testing, maturation, selection and experimental
mortality can be well controlled with this design.

Interrupted time series design

In interrupted time series design changes in behavior as a function of some naturally occurring event
(such as a natural disaster or the introduction of a new law) is measured rather than by manipulation
of an independent variable. The naturally occurring event is a quasi-independent variable, and
comparisons of behavior prior to and after the treatment are observed.

O1 O2 O3 O4 treatment O5 O6 O7 O8

Time

Equivalent Time Samples Design

In equivalent time samples design a single group is used and the treatment is administered to the
group repeatedly in some systematic way. The treatment is introduced and then observations are
made after that observations are made without the treatment and this sequence is repeated. This
design is most appropriate when the effects of the treatment are temporary or transient (Campbell
and Stanley, 1963). Extraneous variables like maturation, testing, regulation, selection, experimental
morality and instrumentation can be controlled with this design.

Treatment O1 no treatment O2 treatment O3 no treatment O4

Time

Nonequivalent Control Group Design

Sometime reconstitution of subject is not allowed and they cannot be randomly assigned to the
control group and the experimental group. In nonequivalent control group design a comparable
group of subjects is selected and observed for the same time period as the group that receives the
treatment however the control group is not equivalent because it comes from a different
community. Even though this design is helpful in making comparison there are some disadvantages.
The validity of the design is compromised if the two groups differ on some important variable
(Campbell and Stanley, 1963). Second if either group is selected on the basis of extreme scores on
the pretest then any shift of scores from the pretest to posttest towards the less extreme values may
be due to regression towards the mean rather than to the effect of treatment.

Group 1: O 1 O2 O3 treatment O4 O5 O6

Group 2: O 1 O2 O3 O4 O5 O6

Time

Counterbalanced Designs

In counterbalanced designs experimental control is achieved by random application of experimental

treatments is used for experimental control. In counterbalancing, the various treatments of the
experiment are assigned in a different order for different subjects. Such designs are called Crossover
designs (Cochran & Cox, 1957; Cox, 1958), Switch-over designs (Kempthorne, 1952), and Rotation
experiments by McCall (1923) Underwood (1949) used the name counterbalanced designs first.
Usually, the Latin-square management is used in this design, where each treatment appears only
once in each column and in each row.

A counterbalanced design with 4 groups of subjects and 4 treatments will look like:

Group A: X1O X2O X3O X4O

Group B: X2O X4O X1O X3O

Group C: X3O X1O X4O X2O

Group D: X4O X3O X2O X1O

Extraneous variables like maturation, testing, regulation, selection, experimental morality and
instrumentation can be controlled with this design.

Separate Sample Pretest-Posttest Design

This design is best when subjects cannot be assigned to treatment at a time. A sample is selected to
administer the treatment and then again another sample is selected for administering the same
treatment again hear the the subject cannot be assigned to treatment or not treatment conditions.
Since no true experimental design can be applied here a separate sample pretest-posttest design
may be useful. In this design variables like history cannot be controlled if the design is used without
repetition. Maturation and interaction of selection are also some variables that cannot be
controlled.

O1 treatment O2

O3 treatment O4

Patched-Up Design

In patched-up design the experiment is started with an inadequate design and some features are
continuously added to it so that the factors producing in validity may be maximally controlled.
Usually the combinations of the inadequate designs are used so that their weaknesses are minimised
and their strengths are increased. In this type of design the experimental control when and to whom
the treatment is given.

Longitudinal study

In longitudinal design a , a single group of participants is measured to observe the effect of passage
of time. In this design the effects of extraneous events that occur during the study are emphasized
and they may not generalize over time. Longitudinal study may also be done for a shorter period of
time.

Longitudinal study keeps subject variables reasonably constant between the conditions because the
same group is observed. It also avoids problems associated with sample non-equivalence, and helps
in the formulation of cause-and-effect statements more accurately than other research designs.
Growth increments and patterns can be studied.

However, it lacks temporal validity as lengthy longitudinal studies may not generalize to future
generations because the society and culture are regularly and constantly changing with time. This
study may be subject to carryover effects. Environment of the subject and unnecessary extraneous
variables cannot be controlled and this kind of design is expensive and time consuming. Other
problems of a longitudinal design include participant mortality, testing effects, and the time needed
to collect even small amounts of data.
The Cross-Sectional Design

Cross-sectional design is a between-subjects quasi-experiment in which several participants from

each of a number of age groups are selected and observed at different points in temporal sequence.
Groups are created on the basis of chronological ages of the participants at the time of the study.

Cross-sectional study can be conducted rather quickly and easily and is less expensive than
longitudinal study. The cross-sectional design allows the investigator to obtain useful developmental
data in a relatively short period of time.

Generation effects are a major problem when you use a cross-sectional design to evaluate age-
related changes in behavior of participants, it refers to the influence of generational differences in
experience, which become confounded with the effects of age. This threatens the internal validity of
cross-sectional studies. Participants of the same chronological age but different maturational age are
grouped together. This design does not indicate the direction of change taking place in the groups. It
also ignores the continuity of development. The comparability of the groups being studied is always
uncertain in cross-sectional design.

Cohort Design

In cohort design longitudinal study is conducted on several groups from different generations.
According to Schaie (1965) the cohort-sequential design combines the two developmental designs
and allows to evaluate the degree of contribution made by factors such as generation effects. The
design includes features of both the cross-sectional and longitudinal designs. Each group is known as
a cohort group and the differences between the groups are known as cohort effects.

CONTROL OF ER

Extraneous variables are those that operate in the experimental situation in addition to the
independent variables and affect the dependent variables. It is, therefore, essential that extraneous
variables must be controlled. Extraneous variables can threaten the internal and external validity of
the study.

Five important ways to control the extraneous variables:

(1) Technique of elimination

(2) Constancy of conditions

(3) Balancing

(4) Counterbalancing

(5) Randomization

Technique of Elimination

The easiest and simplest way to control the extraneous variable is to eliminate it completely from
the experimental situation. Using a sound-proof room is an example of elimination of sound.
However, this technique does not work for the majority of variables used in experimental situations.
Extraneous variables like intelligence cannot be controlled through this technique.

Constancy of Conditions

The second method is to hold the values extraneous variables values constant for all participants in
all conditions. This is known as constancy of conditions. Holding constant the extraneous variables
like instructions given to participants, and conditions like the level of illumination, room temperature
is a good way to control their effects on the experiment. The apparatus used should also be held
constant in all conditions for all participants. In some experiments it is essential to hold the
organismic variables like sex, age, intelligence constant which become important extraneous
variables. To accomplish this, the researcher usually chooses participants of same age and sex.

Balancing

The technique of balancing is used in controlling the extraneous variables when the researcher is
either unable to identify the external variable or attempt to actively control them through balancing.
In the first situation equal treatment is given to all subjects but the experimental group receives the
experimental treatment and the control group doesn't. The extraneous variable therefore influences
both the experimental group and the control group in equal manner and their effect is balanced.
Show the changes occurring in the behavioral measure for the dependent variable can be attributed
to the independent variable. The effects of extraneous variable can be balanced out by allowing
them to operate on the control group as well.

In the next situation the experimental clearly knows which extraneous variable he wants to control.
For example to control the effect of sex by balancing the experimenter may assign an equal number
of subjects from each sex to each group. This will balance out the effect of sex upon dependent
variables. In balancing is that the participants in the control group and the experimental group are
initially equivalent. They are only treated differently.

Counterbalancing

The technique of counterbalancing is used to control extraneous variables like the practice effect
and the fatigue effect. In some experiments each subject is required to serve under two or more
different experimental conditions and there is a probability that the participants performance may
improve due to practice effect on may decrease due to factors affect these are together known as
the order effects.

Therefore an encounter balancing each condition must be presented to each subject an equal
number of times and each condition must occur an equal number of times at each practice session.
Each condition must precede and follow all other conditions an equal number of time. Counter
balancing maybe of two types- complete counterbalancing and partial counterbalancing.

Counterbalancing is used to control asymmetrical transfer or differential transfer. Asymmetrical

transfer occurs when the transfer from condition A (when occurs first) to condition B is different
from the transfer from condition B (when occurs first) to condition A.

Randomization

Randomisation is the best technique to control extraneous variable. It refers to a technique in which
each member of the population are universe pageant equal and independent chance of being
selected. When other techniques cannot be used or when the extraneous variable cannot be
specified this randomization is used. In randomisation extraneous effecst are randomly distributed
over the experimental and control conditions.

In this technique, treatments are randomly assigned to the experimental groups. It is assumed that
the extraneous factors are present equally in all the groups. It may also be used for the selected of
participants where randomization ensures that every member an equal chance of being assigned to
any group.

SINGLE SUBJECT DESIGN

Withdrawal designs
A-B DESIGN
The A-B design is the basic single-subject design. It is made up of baseline for a behaviour A and the
intervention (B). It is predicted that the behaviour would continue to exist if no treatment is
administered. If the intervention causes any behavioural change in the the prediction it may be
attributed to the effect of the treatment. A-B design provides evidence of an association between
the intervention and the change

The logic of the design is that the behaviour would continue in the same way as a baseline if the
intervention is not introduced. If the intervention changes the behaviour of the response measure
increases then and it can be concluded that the intervention produces the change. However it
cannot be certainly said whether any response change was produced because of that specific
intervention. It might have changed due to something else. This is called the placebo effect which is
one of the limitations of this design. From this design, no causal statements can be made.

A-B-A DESIGN
The A-B-A design builds on the A-B design by integrating a posttreatment follow-up that would
include repeated measures. This design consists of a baseline (A), an intervention (B) and a baseline
(A) in this sequence. These three phases represent a series of measurement. The behaviour is first
study to examine whether it changes from A to B the treatment condition and if it comes back to
baseline if the treatment is withdrawn. If the behaviour increases during the treatment and then
decreases when the treatment or intervention is withdrawn and comes to baseline it can be
concluded that the response change is a function of manipulation of the independent variable or
intervention.

The logic is that if the intervention is effective, the target problem should be improved only during
the course of intervention and the target scores should worsen when the intervention is removed.
This design is more powerful and yields more reliable and valid data. A-B-A design provides
additional support for the effectiveness of an intervention.

A-B-A-B DESIGN
A further extension of A-B design is the ABAB design, in which the intervention is re-implemented in
a second “B” phase. ABAB designs have the benefit of an additional demonstration of experimental
control with the reimplementation of the intervention. This design is comparable with the
equivalent time samples design. The baseline and interventions are repeated twice. The logic is that
the behaviour may change from the first baseline (A) to intervention (B) and may decrease following
withdrawal of (B) which is the second (A) and finally the behaviour again increase due to the
introduction of the second intervention (B). The functional relationship between the intervention (B)
and the behavioral measures is strengthened as a result.

This design is more useful for examining the intervention effect than the other withdrawal designs.
One common variation of this design is the A-B-A-C-A design. This is used when effects of two
treatments B and C relative to the baseline has to be implemented. Here the researcher can state
whether B or C affect behaviour, but he cannot tell whether one treatment is more effective than
the other.

ALTERNATING TREATMENTS DESIGN

In this design A and B are the two different treatments that are alternated randomly with a single
subject. The treatment A is withdrawn and replaced not by the baseline but by another treatment,
that is, B. The basic purpose of this design is to evaluate the relative effectiveness of two or more
than two treatments.

Reversal designs
2. Reversal Designs

In this design usually two alternative incompatible behaviours are chosen and the researcher
establishes the baseline for each behaviour. One behaviour is subjected to one type of experimental
treatment and the other alternative and incompatible behaviour is subjected to another type of
experimental treatment. When sufficient data has been collected the treatment conditions are
reversed so that the first behaviour receives the treatment initially given to the second behaviour
and the second behaviour receives the treatment initially given to the first behaviour. Thus virtually
there is reversal of the treatment for the two behaviours.
Multiple baselines design
Multiple Baseline Designs The multiple baseline design is the most common used experimental
design with regard to evaluating intervention effectiveness. It is a powerful strategy that allows
researchers to examine the impact of an independent variable through multiple settings, behaviors,
and participants without resorting to removing the intervention so that it verifies the development
in behavior (Cooper et al., 2007). As discussed previously in reversal design, the nature of reversal
design requires the removal of the treatment in order to verify the prediction that was established in
the baseline condition. However, the equation is different in the multiple baseline experimental
design. There are three basic types of multiple baseline designs:

• Multiple baseline across behaviors, that is, it encompasses two or more different treatments of the
same participant or participants.

• Multiple baseline across settings, that is, there is the same treatment with the same participant in
different settings, situations, or time periods.

• Multiple baseline across subject, that is, it encompasses the same treatment with two or more
participants (Cooper et al., 2007).

In a multiple baseline design, researchers combine a baseline condition with a treatment condition
across participants, behaviors, or settings. A stable baseline is obtained for a participant, behavior, or
setting, then the treatment is implemented for the first participant, behavior, or setting. The data
are gathered for another participant, behavior, or setting. When the alteration occurs for the
baseline that receives the treatment, the procedure is repeated with the second behavior,
participant, or setting (Tankersley et al., 2008).

OR
The basic format is a concurrent multiple baseline design, in which a series of A-B designs (although
A-B-A or A-B-A-B designs could also be used) are implemented at the same time for at least three
cases (clients, target problems, or settings). Therefore, the data are collected at the same time. The
unique feature of this design is that the length of the baseline phase is staggered (see Exhibit 7.12)
to control for external events (i.e., history) across the three cases. The baseline phase for the second
case extends until the intervention data points for the first case become more or less stable.
Similarly, the intervention for the third case does not begin until the data points in the intervention
phase for the second case become stable. The second and third cases act as a control for external
events in the first case, and the third case acts as a control for the second case.

As a research method, multiple baseline designs are particularly useful. They introduce two
replications so that if consistent results are found, the likelihood that some external event is causing
the change is reduced. If some extraneous event might impact all three cases, the effect of the event
may be picked up by the control cases.
Across subjects. When a multiple baseline is used across subjects, each subject receives the same
intervention sequentially to address the same target problem.

Across target problems. In this case there is one client, and the same intervention is applied to
different but related problems or behaviors. The application of the intervention as it relates to the
target problems or behaviors is staggered.

Multiple baseline designs can be applied to test the effect of an intervention as it is applied to one
client, dealing with one behavior but sequentially applied as the client moves to different settings.

ADVANTAGES OF SINGLE SUBJECT DESIGN

There are some advantages or strengths of single-subject experimental research. As such, this
research is commonly conducted in psychology. Some of the main advantages are as follows:

1) The biggest advantage of single-subject research design is its ability to carry out a scientific
investigation with only one subject (or sometimes two).

2) This advantage is more important for a psychologist than for other social scientists, because it
saves much time in dealing with many subjects or large N research or between group researches and
he is able to have full concentration on only one subject

3) Single-subject research design allows the researcher to control the experimental situation more
effectively by establishing a good, obvious and continuous measure of the dependent variable (DV)
throughout the experimental situation.

4) Such researches rarely require statistical tests to be performed for evaluation of data. Only
through visual inspection the experimenter examines such factors as changes in the magnitude and
rate of behaviours being studied (Kazdin, 1982). This point is a big advantage for those who dislike
carrying out statistical computations

5) Both exploratory and descriptive researches can easily be carried out with single-subject design.

6) Single-subject research design allows the researcher to eliminate and hold constant extraneous
variables that do not show up until after the investigation is under way. In fact, intra-subject
comparison in case of single-subject research provides better control of extraneous variables than
the intracomparison, which is the case with large N experimental researches.

7) Single-subject designs are preferred because they are highly flexible and highlight individual
differences in response to intervention effects. In general, single-subject designs have been shown to
reduce interpretation bias for counselors when doing therapy.

DISADVANTAGES OF SINGLE-SUBJECT DESIGN

Research designs are traditionally preplanned so that most of the details about to whom and when
the intervention will be introduced are decided prior to the beginning of the study. However, in
single-subject designs, these decisions are often made as the data are collected. In addition, there
are no widely agreed upon rules for altering phases, so, this could lead to conflicting ideas as to how
a research experiment should be conducted in single-subject design. The major criticism of single-
subject designs are:

1) Carry-over effects: results from the previous phase carry-over into the next phase.

2) Order effects: the ordering (sequence) of the intervention or treatment affects what results

3) Irreversibility: in some withdrawal designs, once a change in the independent variable occurs, the
dependent variable is affected this cannot be undone by simply removing the independent variable.

4) Ethical problems: Withdrawal of treatment in the withdrawal design can at times present ethical
and feasibility problems.

5) Practical limitations: single subject research is time consuming research. It generally takes several
fortnights or months to complete whereas much large end research design can be carried out in only
one session.

6) Baseline Problems: Barlow & Herson (1984) have discussed many problems related to baselines in
single subject research. One of the basic assumptions of single subject research is that stable
baseline has been established. If baselines varies it is difficult for the researcher to conclude a
reliable change in behaviour following treatment or intervention.

7) Researcher’s Bias: Some biases on the part of a researcher are often seen on introducing the
intervention or treatment. Generally the researcher waits for some cues indicating change in
subject’s Behaviour and then he introduces a withdrawal of the treatment. Such practices by the
researcher are not scientific.

8) Lack of External Validity: Single-subject research in comparison to the between groups research or
large N research has poor degree of external validity or generalisability. Since such an experiment is
conducted only on one subject the conclusions obtained are difficult to be generalised to other
participants.

Factorial designs

A factorial design is one in which two or more variable or factors are employed in such a way that all
the possible combinations of selected values of each variable are used (Mcburney & White, 2007).
According to Singh (1998), Factorial design is a design in which selected values of two or more
independent variables are manipulated in all possible combinations so that their independent as
well as interactive effect upon the dependent variable may be studied.

Main effect is the simplest effect of a factor on a dependent variable. According to Mcburney &
White (2007) main effect in a factorial experiment, the effect of one independent variable, averaged
over all levels of another independent variable.

An interaction is the variation among the difference between mean for different levels of one factor
over different levels of the other factor. According to Mcburney & White (2007) Interaction means
when the effect of one independent variable depends on the level of another independent variable

1. The experimenter can manipulate two or more independent variables all possible combinations.
2. In factorial design subgroups are subject must serve under all possible combinations of the levels
of the factors or independent variable.

3. Factorial design allows studying independent effect and the interactive effect of 2 are more
independent variables.

ADVANTAGE OF FACTORIAL DESIGN

1. Factorial design enables the researcher to manipulate and control two or more independent
variables simultaneously.

2. This design enables researcher to study the separate and combined effect of number of
independent variables. This interaction effect cannot be studied in single IV designs.

3. Factorial design is more precise than single factor design (Kerlinger, 2007).

4. By factorial design it is possible to find out the independent or main effect of independent
variables and interactive effect of two or more independent variables. The experimental results of a
factorial experiment are more comprehensive and can be generalised to a wider range due to the
manipulation of several independent variables is one experiment

LIMITATION OF FACTORIAL DESIGN

1. When there are more than three independent variables each with three or more levels are to be
manipulated together, the experimental setup and statistical analysis become very complicated.

2. In factorial experiments when the number of treatment combinations or treatments becomes

large, it becomes difficult for the experimenter to select a homogeneous group.

3. Sometimes the treatment conditions that result from manipulation of several variables are
meaningless and not necessary for the study. This results in waste of time and effort.