Case Studies & Experimentation (Internal validity- whether the…
Case Studies & Experimentation
A case study is defined as an empirical inquiry that investigates a contemporary phenomenon within its real-life context and in which multiple sources of evidence are used.
Interviews for case studies are often informal and unstructured discussions with a key informant of the case. Using semi-structured interviews, the researcher can get the informant's perspective on the issue and get confirmation of the insights and information the researcher already holds.
A single case study may be justified if used as a closing critical study or investigating extreme or unique cases or pragmatic reasons.
When using multiple case studies, there needs to be replication logic- the outcomes should be similar across studies. the researcher can compensate weaknesses of one approach with the strength of another approach.
emphasizes embedding the phenomenon in its real-life context. allows detection of unexpected patterns and potential explanations. The results are often not generalizable to a population, they are generalizable to a theoretical disposition.
Case studies are more appropriate if the number of variables that needs to be considered is quite large with an important context.
Experiments involve intervention by the researcher ; manipulating an independent and explanatory variable in a setting and observe how it affects the subjects being studied and whether the hypothesized dependent variable is affected by the intervention.
The researcher is able to manipulate the IV. Use of a control group serves as a comparison to assess the existence and potency of the manipulation and pre- and post-test measurements allow checking that the manipulation occurred before the outcome.
Contamination from extraneous variables can be controlled more effectively that other research designs, helping the researcher to isolate experimental variables and evaluate their impact over time.
The experimenter can schedule data collection and have the flexibility to adjust variables and conditions that evoke extremes not observed under normal circumstances. In addition, the experimenter can assemble combination of variables for testing rather than having to search for their accidental or chance appearance in the study environment.
Replication of the experiment with different subject groups and conditions leads to the discovery of an average effect of the IV across people, situations and times.
The artificial environment. It may also be costly, outrun the budget of other primary data-collection methods.
Generalisation from non-probability samples can pose problems- the extend to which a study can be generalised is open to question. Volunteer subjects are often those with the most interest in the topic.
The number of variables that one can include is limited than a survey. Experiments are not appropriate for research problems that involve many influential factors.
In experimentation, problems investigated need to be of the present or immediate future and studies about intentions or predictions are difficult. Factors included in the experiment need to be easy to manipulate.
There are limits to the types of manipulation and control of people that are ethical.
basis for conclusion are:
1) The IV and DV are correlated.
2) Time order- the DV should not precede (come before) the IV, although they may occur simultaneously.
3) No other extraneous variable influenced the dependent variable.
Steps of experimentation
1) Select the relevant variable- select variables that are the best operational representations of the original concepts. determine how many variables to test and the number of subjects being tested. Then we need to select and design appropriate measures for the variables.
2) Specify the level of treatment- The levels assigned to an IV should be simplistic and common sense. Consider using a control group.
3) Control the experimental environment- When subjects do not know if they are receiving the experimental treatment, they are said to be blind (to control subject reactions to expected conditions).
When the experimenters do not know if they are giving the treatment to the experimental group or the control group, the experiment is said to be double blind (to control experimenter influence).
4) Choose the experimental design- Careful selection strengthens the generalisability of results beyond the experimental setting.
5) Select and assign subjects- Must be representative of the population. Random assignment to the experimental group is required to make the groups as comparable as possible with regard to the DV. matching may be used; employing a non-probability quota sampling approach to have subjects in both groups be similar to each other.
6) Pilot-test, revise and test- reveal errors in the design and improper control of extraneous or environmental conditions.
Internal validity- whether the conclusions
we draw about a relationship
truly imply cause?
History- Between the control measurement of the DV (done before the manipulation of the IV) and the after-manipulation measurement of the DV, many events could occur to confound the effects of the treatment.
Maturation- When a study covers a long period, a subject can become hungry, bored or tired and other changes within the subject through the passage of time may affect response results.
Testing- The process of taking a test can affect scores of a second test (learning effect) as participants may take additional efforts to learn more about the topic.
Instrumentation- Using different questions, observers, interviewers and sometimes using the same observer for all measurements may be problematic. Observer experience, boredom, fatigue and anticipation of results can distort results of separate observations.
Selection- of subjects for experimentation and control group. The groups should be equivalent in both groups.
Statistical regression to the mean- random fluctuation over time
Experiment mortality- attrition of subjects causes the groups to change, higher chance of attrition for experimental group.
Diffusion or imitation of treatment- if group members talk to each other about the treatment, eliminating the differences between the groups.
External validity- does the observed causal relationship generalize across
people, settings and times?
The reactivity of testing on X- when subjects are sensitised via a pre-test causing them to respond to the stimulus (X) in a different way.
Interaction of selection and X- process of selection of subjects affects ability to generalise.
Field experiments occurs within a natural setting where participants do not know that their behaviour is being monitored. It may be more representative of the population and subjects are not aware of their participation, therefore, less likely to conceal or adjust their behaviour. There is no experimenter effect that may occur in a laboratory. However, the researcher has fewer opportunities to control the research setting variable and may raise ethical questions as participants do not give prior consent to participate in the research.