Hello I am trying to run a moderation analysis but will need to use G*Power to determine my sample size. Just wondering if anyone could assist me with the following:
1) the parameters to set
2) Also what effect size/power should I use?
3) If my outcome measure is pre vs post test change on a questionnaire, would Hayes process macro be a good program to use or should I use SPSS instead?
IV: Intervention
M: Language (3 lvls)
DV: Pre test vs post test of an outcome measure
Thank you.
Hi,
To determine the sample size for moderation with GPower, select 'F tests', then 'Linear multiple regression, fixed model, R² increase' and 'A priori' power analysis. Input effect size, power (usually 0.80+) and alpha error probability (often 0.05). GPower will give you the needed sample size. It seems Hayes PROCESS might be a macro for SPSS, not separate software. Is SPSS your planned tool for moderation analysis?
Hope this helps.
In my personal opinion, G*Power is not very helpful for moderattion analyses. It is very hard to determine, how to interpret the effect size of interest or how much addtional variance is really of interest and matters. It is simpler to think in the measures of the dependent variable, i.e. the regression coeffcients, or more specifically, the coefficient for the interaction. How much change in the focal predictor, conditional on the moderator, is of practical relevance? It becomes simpler, if you would simulate different scenarios instead of using G*Power. Artificial recommendations for small, medium or large effect sizes are not really helpful, since these categories heavily depend on the research area and are not natural laws.
Just my two cents....
Rainer Duesing has given you some excellent advice. Re his comment about "artificial recommendations for small, medium or large effect sizes", see also this 2020 opinion piece by Correll et al.
https://www.sciencedirect.com/science/article/abs/pii/S1364661319302979
Article Avoid Cohen’s ‘Small’, ‘Medium’, and ‘Large’ for Power Analysis
Hello Hsin-Yuan Chen
thank you so much for your suggestion on G*Power. I was planning to use Hayes Process Macro in SPSS but realised Process may not be able to run my analysis since my outcome measure involves repeated measure (pre vs post outcome measure). Hence I was wondering if there was another analysis I should use instead:
1) Linear Mixed Models (not too familiar with this)
2) Compute a percentage change for my outcome measure (pre vs post) then run moderation on Hayes as per normal with Model 1.
Hello Rainer Duesing and Bruce Weaver
thank you for sharing your perspectives. I think I have only been introduced to G*Power. But it was interesting to hear this perspective. Thank you Bruce for the reading resource as well which I will definitely read.
Hi,
To determine the sample size for moderation with GPower, select 'F tests', then fixed model, R² increase' and 'A priori' power analysis. Input effect size, power (usually 0.80+) and alpha error probability (often 0.05). GPower will give you the needed sample size. It seems Hayes PROCESS might be a macro for SPSS, Is SPSS your planned tool for moderation analysis?
Hope this helps.
Venetta Ho
Hi,
For your study with pre-post test changes, using LMM in SPSS is appropriate as it effectively handles repeated measures. Set up your model with your intervention and language as fixed effects. Alternatively, calculate the percentage change between pre-post test and run a standard moderation analysis using Hayes' Model 1.
Hope this helps.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Graphs