Switch branches/tags
Nothing to show
Find file
Fetching contributors…
Cannot retrieve contributors at this time
81 lines (78 sloc) 4.36 KB
The Problems with Small N Designs
* Lack of generalizability
* Can we insure generalizability in correlation research?
* Too uncontrolled/too many threates to internal validity
* There are ways to establish validity... see next slides
* Takes too long to conduct
* Sometimes in depth, longitudinal research is needed to really understand something.
Establishing Validity in Small N Designs
* Kratchowill, Mott, & Dodson (1984) criteria for establishing validating small N studies
Measurement Criteria
* Use objective data instead of subjective data
* Use more than one dependent variable
* Use multiple sources of information
* Frequent Assessment
Replication Criteria
* The more a research finding is replicated the more people can believe in it. It increases generalizability and identifies limits of theories
* Yin (1994) has a model in which several case studies are done and the results are compared across them to see if they all converge.
Control Criteria
* There are many threats to internal validity in case studies: History, Maturation, Regression towards the mean.
* Yin (1994)
* Test Cases: Like an experiment, when the IV is present and it has an effect on the DV.
* Control Case: What happens to the D in the absence of the IV.
* Sometimes, especially in clinical settings, this is not always the best way to go, as you may end up withholding treatment
Impact Criteria
* According to Kazdin (1997), the larger the magnitude of a treatment on a DV (i.e., effect size), the more certain the researcher can be that history, maturation, and regression are not playing a role.
* Assessing Impact (Mostly Clinical)
* Chronicity
* Can help rule out maturation
* Large Magnitude
* Can help rule out regression to the mean
* Immediate Impact
* History will be less a factor if the impact of the treatment can be seen immediately following its implementation
* Follow Ups
* If the effect lasts at follow ups we somewhat confident that there is no placebo and regression effects
Treatment Criteria
* Having control over when the treatment occurs
* Standardized they occur the same way for every person
Single Case Studies
* Terminology:
* A: Usually refers to the baseline condition
* Thing of this like a control session in a within-Ss design
* B: The treatment conditions — The IV
* C: A second treatment condition — A Second IV
A — B Designs
* Simplest design
* Not that often used because:
* It is more suscpectial to threats of internal validity such as regreession to the mean, history, and maturation
A — B — A Design
* Logic: The IV should bring about a change in the DV. When it is removed, the DV should return to baseline. This is called a reversal
* Sometimes also called a withdraw design
Multiple Baseline Designs
* In these designs, researchers create or establish multiple dependent measures.
* Then treatments are established at different times
* We can rule out history by showing that changes occur only when the treatment happens
The Changing Criterion Design
* It is nice when we can see large changes in the behavior of interest as a function of our treatment
* Not all treatments work in this way. Somethings change gradually
* Desensitization therapy — often requires multiple sessions to notice a change in the dependent measure.
Case Studies
* An in-depth, usually long-term examination of a single instance of a phenomenon.
* Choosing cases to study
* If you can, select a situation in which you might be able to manipulate an IV.
* Select cases that you can easily access and will have access to for large chunks of time
* Data Collection
* You only get one shot at it, so be careful
* Carefully select the DVs, when data collection will occur, where there are sources of data
* Also be ready for changes in the procedures and be able to adapt accordingly.
* Search for disconfirming evidence and make sure to report it, if you find any
* Make sure to keep and establish a chain of evidence
* Needed to effectively support your conclusions
Chain of Evidence
* Research report should state where the data from each conclusion comes from
* Reports should include transcripts of interviews, etc.
* Report should state how each piece of data was collected (e.g., interview)
* Report should mention what measures or pieces of data were specifically used to answer certain questions.
Data Analysis for Case Studies...
* Qualitative data analysis
* No Stats