Professional Documents
Culture Documents
Research Designs
Chapter Summary
The research design process starts with determining the research questions. The
chapter opens with an excerpt from the Carrolls Alices Adventures in Wonderland
story. We see that some direction is required to start a journey. Whether it be an
actual destination or to simply explore and experience the journey, both are
directed in some way. Conducting social research can very much be seen as a
journey.
The intent of research may be to try to describe connections between variables,
exploring a social phenomenon, how it changes over time, the meanings that people
attach to things in their lives, to develop generalizations that apply to broad
categories of people. Further you need to consider the type of explanation you
would like to use.
Quantitative research relies on cause and effect based in laws and principles that
can be extrapolated to people both inside and outside of the study. This is referred
to as nomothetic explanation. To be acceptable it requires the presence of three causes
or conditions.
1. Correlation (An affected variable changes with the variable that causes it)
2. Time Order( The cause must occur prior to the effect)
3. Non-spurious (There is no alternate explanation for the correlation)
Qualitative research is typically ideographic. It tries to get deep understanding of the
perceptions and feelings of research subjects. Typically it only applies to localized
groups/activities in what is called an idiographic explanation. It addresses only the
activity of those involved in the research. The data is not meant to be extrapolated
to a large external social group.
Methods used in qualitative research range from non-invasive self-completed questionnaires through to participant observations which may entail infiltration of social
groups. Case studies for example examine a specific place, time, event, person,
community, organization. The researcher must decide how best to get access to data
that will explain the particular case.
Causality
Causality in social research is directly related to variables. Variables quite simply are
characteristics or attributes of people or things that are different or change (vary)
within the population. Examples may include determined physiological attributes
such as gender, athletic ability, and/or subjective characteristics such as interest in a
particular topic, beliefs.
There are three key criteria for evaluating a research project:
1. Reliability Would the same results occur if the research was repeated again
with the same research subjects? This speaks to the impact of differing
locations, timing and the researcher on the outcome.
2. Replicability Would the same results occur if the research was repeated by
other researchers? This speaks to good research procedures.
3. Validity Is there integrity in the outcomes and conclusions? There are
three types of validity.
i) Measurement or construct validity
Are you actually measuring what you what to measure?
ii) Internal Validity
How confident can you be that the variables are co-related? Is the
secondary or dependent variable actually impacted by what you think
is the primary or independent variable.
Independent variable proposed cause (occurs first)
Dependent variable proposed effect (occurs second as a result of the
specified independent variable that occurred/existed before it)
iii) External Validity
Is the research design or setting too artificial? Do the study findings
relate to the real world that exists beyond the particular research
environment? Qualitative and quantitative methodologies use different
methods to overcome this concern.
Qualitative methodology incorporates naturalistic observations.
Quantitative methodology uses representative sampling to overcome
this concern and generalize the findings to extrapolate them to a larger
population.
Lincoln and Guba (1995) argue for different standards to judge qualitative methods.
They argue for the inclusion of trustworthiness. Trustworthiness, however, simply
includes four principles that mirror the measuring criterion for quantitative
methods.
Trustworthiness
Quantitative Criterion
Credibility
Internal Validity
Transferability
External validity
Dependability
Reliability
Confirmability
Replicability
Social Research Methods, Third Canadian Edition
Oxford University Press Canada, 2012
Experimental Design
Experimental design is common in psychology and organizational studies because of
the ability to manipulate independent variables to determine the influence on the
dependent variable. In sociology and political science the process of variable
manipulation is much more difficult, if not impossible to do. Many variables (i.e.,
poverty, war) have long-term issues that cannot be set up as experiments and ethical
concerns prevent experiments. See the Stanford Prison Experiment listed below.
Even where experimental design may seem operationally sound for qualitative
research, it will not get at the perceptions and feelings of the research subjects that
is a key component of qualitative research. The one may lead to the other. Further
there is a difference between laboratory and field experiments. Laboratory experiments occur in artificial settings (e.g., candid camera) whereas field experiments
take place in real-life surroundings. Because laboratory experiments take place in
artificial environments it is easier to control the research environment, easier to
randomly assign research subjects, which enhances internal validity, and it is easier
to replicate.
Experimental design requires some specific components to be valid:
A group that is essentially left alone to function as it did prior to the test is
required. It is called the control group. It provides a baseline to which
researchers will compare the experimental, or treatment, group.
The experimental (treatment) group receives some sort of treatment or manipulation.
Research subjects must be divided between the control and experimental
groups in a random fashion. This allows the researcher to
o presume the two groups are the same, relative to the independent
variable; and
o accept any change in the treatment group as an occurrence that is related
to the treatment received.
In the classic experimental design the independent and dependent variables are
identified. The dependent variable is observed or measured (pre-test) in each of the
control and treatment groups and recorded as T1 (time 1). The treatment group
receives the treatment/manipulation while the control group is left alone. The
dependent variable is observed or measured (post-test) in each of the control and
treatment groups and recoded as occurring at T2 (time 2). Any changes in each
group are noted. Ideally change will only occur in the treatment group. See
Rosenthal and Jacobson below.
True experimental evidence would eliminate all other possible (rival) explanations
for the change in the dependent variable in the treatment group. But most social
experience involves several and complex issues. So the use of a control group, and
the random assignment of research subjects between the control and treatment
groups, is aimed at increasing internal validity.
Social Research Methods, Third Canadian Edition
Oxford University Press Canada, 2012
Quasi-experimental Design
Quasi-experiments have some characteristics of the experimental model but lack
one or more of the internal validity requirements. Natural experiments for example
occur when there some change to a group by people who are not researchers, but
there is data available for researchers to collect and analyse for its effect.
Cross-sectional Design
Cross-sectional design is the gathering of data at a specific point in time. There is no
manipulation of the independent variable or treatment group. The data gathered is
meaningful because two or more measurable variables are compared for patterns
of association. Methods used may include questionnaires, structured interviewing,
and structured observation. This design tends to be used in quantitative studies
because the large sample size provides confidence that the suggested correlation
between identified variables occurs regardless of other potential influences and
because statistical techniques typically need large sample sizes.
Because there is no before and after timeline to assess cause and effect, the crosssectional design tends only toward providing patterns of association rather than
causality. Did one variable cause the other to occur, or is the impact of one variable
to another reciprocal?
The issues of reliability, replicability, and validity remain the same for crosssectional design. Are the measure tools appropriate? Are the procedures described
fully? How are the perceived associations substantiated? Perhaps causation is
substantiated through the use of a secondary research method. Was the sample used
a random one?
There is often no ability to manipulate variables in social research, e.g., gender, age,
racial background, ethnicity. That does not mean that causal inferences are not
possible. Rather, the comparison of a large sample size of people of different ethnic
backgrounds may lead to a strong indicator that the dependent variable is impacted
by ethnicity. Likewise the cross-sectional research approach may allow researchers
to rule out the impact of a given pre-existing variable.
The cross-sectional design is used in both quantitative and qualitative inquiry. In
qualitative research the researcher collects in-depth data on what influences
peoples activities. In-depth data on what people believe influences their behaviour
is a discussion of patterns of association. The sense of influence as perceived by
the research subject, by its very nature, implies causality.
Longitudinal Design
Data is collected at a particular time (T1) just as in the cross-sectional research
design. But it is also gathered again at a later time, and perhaps at even further
times (T2, T3, . . .). This process may occur several times. Although there is no
manipulation of an independent variable, the longitudinal design overcomes the
direction of causation problem found in cross-sectional design. It provides insight
into the time order of change in variables to determine, to some degree, which
variables are dependent and which are independent. The additional time and cost
required to conduct longitudinal design projects hinders the use of this methodology.
There are two types of longitudinal design: the panel study and the cohort study.
The panel study relies on the same set of research subjects. They are studied at time
one (T1) and again at a later time, and perhaps at even further times (T2, T3, . . .).
The cohort study examines an experience that people have and then again at a later
time, studies the same experience, although the research later subjects may be, and
most often are, different from the first group. Both study designs share the same
weaknesses:
Attrition (Panel study) Some research subjects may not be available for
subsequent study. This reduces the representativeness of your sample and
thereby the reliability of your data somewhat. The cause for their unavailability may be directly related to a variable not just a random occurrence.
The researcher simply doesnt know.
Timing When is the appropriate time to do further data collection?
Panel Conditioning Being engaged in a panel study (observed) over an
extended period of time may become a variable in the activity of the research
subjects.
Case Study Design
A case study is quite simply the in-depth study of a single case. The case may be an
individual, a family, an organization, an event, a geographic location, a specific
time, etc. The case study is not aimed at achieving external validity. It provides an
in-depth description of a particular set of circumstances.
The term case study may be seen as an overarching research style. It identifies the
individual incidence of activity to be studied as an object of interest in its own right.
The particular method used to study that particular object of interest is determined
by assessing what methodology will be most appropriate given the case scenario.
The methods used may be quantitative or qualitative. Qualitative methods used for
a case study tend to take an inductive approach whereas quantitative methods tend
to take a deductive approach.
Social Research Methods, Third Canadian Edition
Oxford University Press Canada, 2012
12. Which of the following is NOT considered a threat to the internal validity of
research data?
a) history
b) pre-testing may increase the research subjects skills for the post-test
c) a change in the way a test is conducted
d) random assignment of the research subject to control and treatment
groups
e) research subjects get older
13. Cross-sectional design ________.
a) is the gathering of data at three
b) proves a direct link of causality
c) involves the manipulation of variables at one specific time
d) provides patterns of association
14. Longitudinal design ________.
a) overcomes the direction of causation problem found in cross-sectional
design
b) is considered one of the easiest forms of research design
c) requires data to be collected at least three times
d) may suffer from research subject attrition
15. A case study ________.
a) could be used to support an hypothesis
b) could be used to critique the conclusion drawn from research gathered
through a classic experimental design
c) may provide comparative data from an extreme set of circumstances to
better understand common cases
d) all of the above
True or False Questions
1. Although some say that good research requires some sort of predetermined
direction, it is not always necessary.
True ____
False ____
2. Testing for correlation of variables is the only valid qualitative reason for
conducting social research.
True ____
False ____
3. Internal validity is a confirmation that the dependent variable and the
independent variable are co-related.
True ____
False ____
Social Research Methods, Third Canadian Edition
Oxford University Press Canada, 2012
Media Resources
For a short report of several research projects using different research designs on
the correlation between media and violence, and a comparison of their various
strengths and weaknesses, see
http://psychlotron.org.uk/resources/social/AQA_A2_aggression_mediaantinotes.pd
f
o Does the stated aim of the research affect the conclusion?
o How do you account for the difference in conclusions?
Answer Key
Multiple Choice Questions
1. a (p. 23)
2. d (p. 23)
3. c (p. 24)
4. a (p. 24)
5. d (p. 24)
6. d (p. 24)
7. b (p. 24)
8. a (p. 24)
9. c (p. 25)
10. b (p. 24)
11. d (p. 25)
12. d (p. 28)
13. d (p. 33)
14. a (p. 36)
15. d (p. 39)