Additional research designs 2012-11-15
Quasi-Experiments: similar to a experiment, but does not mean all the requirement as the experiment; also called field-
- Key difference to the real experiment: Less control of extraneous variables.
o Use random assign is missing: groups often naturally occurring
- Some common features of quasi-experiments:
- matching instead of randomization
- time series analysis
- unit of analysis not people—could be an organization, a group or other things
One Group Pre-post—it is very basic;
instances of bullying training for staff instances of bullying
- you start this in a school; it is very hard to isolate the school or staff to only monitor certain types of bullying
so that it would only possible to observe the whole school.
- Problems would happen as: threats to internal validityyou have strong evidence to determine your
dependent variables. But in here we cannot really make the conclusion.
o Threats to Internal Validity
History: event affects study outcomeit not only just meaning the things happen before but
could be present condition
It is whatever happening between the pretest and posttest
Change in subjects over a period of time
Ex: you look at the Gr 8 in the beginning, but ppl are growing olderstaturally
But not only over a long time, saying the beginning to the end of 1 year university
term, ppl will change a lot and understand more about university.
Testing / Repeated testing: they have to do the same test in 3 or more different ways; repeated
test make ppl getting used to the test and may act differently; for example, just filling a
questionnaire may make ppl aware or think differently about their own behaviours according
in bullyingnitions given by the researchers; which may have effect on decreasing or increasing
Subjects dropping outleaving the study; not just because ppl drop out but more
importantly, it is about which types/groups of ppl are dropping; it would not be a
problem if the dropping are random as the sample, but it matters there is particular
group of ppl like ppl are low in suggestibility are more likely drop out a smoking-
Regression to the mean
High or low measurements followed by measurements closer to group mean
you set up to do experiment in a school having high than average bullying in your
pre-test, and later the bullying dropped to the average… but what about the school
was just emerged from two different school, in which case, there always more
conflicts at the first period of emerging…. And this would affect your experiment
o It is like you catch a happy person in a low day;
Non -equivalent control group design
Experimental group instances of bullying training for staff instances of bullying
instances of bullying instances of bullying
Selection bias: preexisting difference between the groups Time Series Analyses
- When you cannot randomize participants, but can get assessments of the DV pre and post-treatment
o Most of the time is the weather of a day;??? Usually you have a control group.
Single-Case Research Designs:
- Use only one case or group to investigate a specific phenomenon.
- Not the same as a case study.
- Uses time-series design.
- Take multiple pre and post-treatment measures.
Advantage of Small-N Design
- Participants from hard to find populations—like it is only one participant for particular activity: ex the person
walking through the Niagara Fall
- Results easy to interpret (often no stats)—looking at the relationship about the height of the hill and the oxygen
level—it is usually very straightforward;
- Can focus on helping one (few) participant(s)
- Problems: end up with the baseline condition
- Effect may not be reversible.
- Ex: You take the sleep pills prescription for sleeping but you did getting better after one month; but you cannot
really 100% sure it is the effect of the pills; the only way to find out is stopping taking the pills
o If you stop taking the pills, and you return to base line; it could be some the residual effect of the pills or
it may be you have new things to disturbing your sleep
o The reason is ppl keeping changing; you cannot give the reason because you don’t have the control on
that to find out it is because of the treatment or because of other issues
o So that you have to go back to the treatment
Multiple Baseline Design
• Testing a treatment effect when effect is irreversible.
• What you need to here is you collect more baseline data
• Baseline data collected on:
• 2 or more behaviours for same individual
• Same behaviour for 2 or more individuals
• Same behaviour across 2 or more situations for the same individual.
Choosing a Research Method
—key thing: all research method have advantage and disadvantages
- The choice is affected by:
• Resources like time and money
• Ethical concerns
• The research question
How does the question guide the choice?
- Description: asking what or how many, getting ideas about why, developing theory
- Case study for unusual cases
o Most often it is guided for cases arrive; it is unusual to happen;
o Ex: nonverbal children, hostage-taking situation, natural disaster
It is hard to set up a experiment to get a situation like this.
But if there is a particular abuse happened, it DID happen and you can use this
- Case study for in-depth examination o Ex: merger of two companies
- Correlational study
o Prediction: if I know X, can I predict Y?
Ex: ppl’s GPA in university and the first year salary after graduate
o Causal: Verifying the why or how
Does drug X lead to the relief of symptom Y?
The major treat off is … you cannot really do int