0todd0000
commented
4 years ago Hello! For categorical variables the Beta values are means. So simply compute 1D means for each subject and each category, then submit these to second-level analysis. For a pre/post design the suitable second-level model is probably a paired t test.
This hierarchical approach implicitly considers within-subject variability. If subjects are highly variable, their means are also expected to be highly variable, implying that variance will be large in second-level analyses. If, however, within-subject variability is small, then the means are relatively stable and are suitable for second-level analysis.
tramarobin
Hi Todd!
I think my design is adapted for random effect analysis as I have a pre/post tests design with multiple trials for each test. I used to calculate the mean of all the trials and perform a paired t-test, but I wish to take into account the within-subject variability.
I have a question about random effect analysis, and more precisely on the level 1 analysis. In the example, linear regressions are used and Beta corresponded regression slopes for each subject. However, in my case, my independant variable is not continuous, but categorical. Can I extract T-continuum (with paired t-test) for each subject as Beta parameter, and conduct the level 2 analysis on Beta ?
Thank you.