0todd0000
commented
1 year ago In general: no, Gpower can not be used to establish sample size for SPM analysis. However, that doesn't necessarily mean that Gpower's results are incorrect.
If trajectory-level analyses like SPM are conducted, then trajectory-level power methods are required to calculate sample sizes. *GPower** does not include trajectory-level calculations so in general cannot be used to calculate power for trajectory-level analyses.
There are several methods and packages for conducting trajectory-level power analyses, for example: power1d. The key difference between trajectory-level power analyses and 0D power analyses (i.e., the type of power analysis conducted in *GPower* is that trajectory-level analyses generally require trajectory effects* (or trajectory-level alternative hypotheses) as demonstrated here.
Thus the similarity between GPower results and trajectory-level power results depends largely on the modeled effect. If you model an alternative effect as a simple pulse at a single location, then the effect is very similar to a single-point effect, in which case trajectory-level results will likely be quite similar to GPower results.
Dear Todd, dear colleagues,
I get the comments from the reviewer:
“ I don’t think a power analysis (G*power) based on a repeated measures ANOVA is applicable to statistical parametric mapping. Statistical parametric mapping is based on random field theory, whereas ANOVAs are based on the general linear model”
*Question: Can we usually perform Gpower analysis establishing the sample size where the SPM approach was used?**
Hope this topic helps to other young researchers :-)
With honor, Lukas