How do you determine whether your sample size is large enough for a mediational analysis (using process)?
Dear Jennifer,
read this paper:
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2843527/
Certainly a subject of much debate. Hayes method is fairly generous with sample sizes ie he does not require the type of enormous sample more traditionally suggested. One justifiable approach is to calculate your required sample using G*power prior to recruitment based on a multiple regression model. That calculation would be contingent upon the strength of effect you hope to show and the number of predictor variables in your model. The paper mentioned by the previous poster is also helpful.
There is this paper as well:
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4334479/
Monte Carlo simulation in general is a very flexible method to do any kind of power calculation. I have never done it to assess mediation models, but others have done this. If you google
simulation of mediation power
the first result is a researchgate thread where there are several codes attached in the responses.
Preacher and Hayes' method is supposedly effective with rather small samples. It is a bootstrapping approach, and is defined as non-parametric with regard to the indirect effect. I found codes for doing it on stata on the web, and the original article has codes for SPSS and SAS, if I remember correctly.
The interpretation will depend on a number of other factors, of course (e.g., was the independent variable experimentally manipulated, etc.).
All the best
Morten
- Badges
- Science topic
- Similar topics
- Psychology
- Developmental Psychology