Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
February 1, 2015
Topic Outline
Nature of, Overview and Classification of Design
Developing an appropriate research design
Experimental research design:
Basic Designs
Types & validity of experimental design
external & internal
February 1, 2015
Learning Outcomes
3
February 1, 2015
RESEARCH DESIGNS:OVERVIEW
4
February 1, 2015
February 1, 2015
Clarity
Relevance
Ease in Analysis and Interpretation
Economy:
February 1, 2015
February 1, 2015
Exploratory
Conclusive
Objective:
Toprovideinsightsand
understanding.
Totestspecifichypotheses
andexaminerelationships
Characteristic
s:
Informationneededisdefined
onlyloosely
Researchprocessisflexibleand
unstructured
Sampleissmallandnonrepresentative
Analysisofprimarydatais
qualitative.Tentative
Informationneededisclearly
defined.
Researchprocessisformal
andstructured
Sampleislargeand
representative.
Dataanalysisisquantitative.
Conclusive
Generallyfollowedbyfurther
exploratoryorconclusiveresearch
Findingsusedasinputinto
decisionmaking
Findings/Resu
lts:
Outcome:
February 1, 2015
February 1, 2015
Focus
February 1, 2015
Descriptive Studies
Descriptions
Descriptions of
of
population
population characteristics
characteristics
Estimates
Estimates of
of frequency
frequency of
of
characteristics
characteristics
Discovery
Discovery of
of associations
associations
among
among variables
variables
Variations
Single/multiple cross- sectional designs
February 1, 2015
Longitudinal studies:
1. The study involves selection of a representative
group as a panel.
2. There are repeated measurement of the
researched variable on this panel over fixed intervals
of time.
3. Once selected the panel composition needs to stay
constant over the study period.
February 1, 2015
Causal Studies
Symmetrical
Symmetrical
Reciprocal
Reciprocal
Asymmetrical
Asymmetrical
Experimental Design
16
An experiment is generally used to infer a causality. In an experiment, a researcher actively manipulates
one or more causal variables and measures their effects on the dependent variable of interest.
Independent variables: Independent variables are also known as explanatory variables or treatments.
The levels of these variables are manipulated (changed) by researchers to measure their effect on the
dependent variable.
Test units: Test units are those entities on which treatments are applied.
Dependent variables: These variables measures the effect of treatments (independent variable) on the
test units.
Experiment: An experiment is executed when the researcher manipulates one or more independent
variables and measures their effect on the dependent variables while controlling the effect of the extraneous
variables.
Extraneous variables: These are the variables other than the independent variables which influence
February 1, 2015
Validity
17
Internal validity: Internal validity tries to examine whether the observed effect
Validity
External validity: External validity refers to the generalization of the results of an
experiment. The concern is whether the result of an experiment can be
generalized beyond the experimental
In most of the experiments the data are collected through sampling process and not
from population.
Factors Affecting External Validity:
The environment at the time of test may be different from the environment of the real
world where these results are to be generalized.
Population used for experimentation of the test may not be similar to the population
where the results of the experiments are to be applied.
Results obtained in a 56 week test may not hold in an application of 12 months.
Treatment at the time of the test may be different from the treatment of the real world.
It is desirable to have an experimental design that has both external and internal
validity.
Randomization
Matching
Use of experimental designs
Statistical control
February 1, 2015
Field conditions
Lab conditions
Simulations
February 1, 2015
addition:
MB
X
MA
The result of interest: (MA MB) i.e. considerable
advantage over After-Only.
Ex. To estimate the effect of price increase on market
share.
Design:
R
R
MB1
MB2
MA1
MA2
Ex. A Firm wishes to test the impact of a P-O-P display. Ten retail stores
selected for inclusion in the treatment group, another ten for the
control group. Sales measured in each group of stores, before and after
the new P-O-P display. The change in sales of the two group is
compared. Controls any initial inequalities between the sales group.
In cases where the interaction is unlikely and control for history and
selection errors is important, the before-after with control group design
is the best design in terms of cost and error control.
February 1, 2015
February 1, 2015
MB1
X
MA2
X
MA4
MA1
MA3
It controls all sources of experimental errors except measurement timing & reactive error
which is not subject to control by designs. No single method of analysis makes use of all
six measurement methods simultaneously
All four groups are preselected in such a way that they are equivalent i.e. they are
selected on a random basis. This means before measurement should be the same in all
four groups except for random variations.
The fourgroup six-study design may be taken as a model for marketing experiments, has
little practical value.
The expense of selecting four groups randomly and making six studies among these
groups make this design impractical for most marketing studies.
February 1, 2015
Participants Perceptions
No deviation perceived
Deviations perceived
as unrelated
Deviations perceived as
researcher-induced
Statistical Designs
30
February 1, 2015
Statistical designs
31
February 1, 2015
Statistical Designs
32
February 1, 2015
Factorial Design
33
February 1, 2015
Factorial Design
Ex. Consider the problem of determining the proper concentration of
sugar & flavor in a soft drink.
One approach may be to make a batch of optimum mixture and have a
sample consumer taste it and indicate and order of preference.
Another approach may be: makeup several batches with different level
of sugar content and flavor constant. Consumer may then taste and
indicate a preference. Later on sugar could be held constant and flavor
varied.
Later approach may indicate that heavy sugar and heavy sugar were
preferred. May not be valid always. The fact may be when the flavor is
strong, the sugar may be less desirable.
So its important to test various levels of sugar content combined with
various levels of flavor
Factorial Design
Suppose four degrees/levels each were selected as possible
Sugar Content
1
A 4.9
B- 6.0
C 5.0
D 3.6
E 6.1
F 7.3
G 5.1
H 3.8
I 8.1
J - 9.2
K 8.3
L 4.6
M 6.2
N 6.4
O 6.2
P 3.2
With the Latin Square design one can control variation in two
directions.
The design requires that extraneous or blocking variables be divided in
to an equal no. of blocks or levels, such as drugstores, supermarkets and
discount stores. The independent variables be divided in to the same no.
of levels, such as high price, medium price & low price.
-Treatments are arranged in rows and columns
-Each row contains every treatment.
-Each column contains every treatment.
-The most common sizes of LS are 5x5 to 8x8
Advantages of the LS Design
1. You can control variation in two directions.
2. Hopefully you increase efficiency as compared to the RBD.
February 1, 2015
February 1, 2015
SUMMARY
February 1, 2015
References:
39
1.
2.
3.
4.
February 1, 2015
Thank you!