Hi FSL experts,
I am relatively new to FSL, and have previously used AFNI. One main difference between packages is that AFNI default is to concatenate the runs together before preprocessing steps and first-level statistics are applied, whereas FSL the default is to model each run separately and then combine them for each subject at level2.
I am analyzing a task designed for a developmental sample so it is very short: 3 runs of 99 TRs. Within each run, there are 24 "task A" trials and 24 "task B" trials. However Task A consists of 3 different emotion conditions (8 trials each). So across all 3 runs, I have 24 trials for each emotion condition for "Task A". My main question of interest involves comparing the emotion conditions for activation and PPI.
I understand that one main advantage of FSL is the capacity to conduct mixed-effects modeling (i.e. level1 by run, level2 by subject, level 3 for group) to better account for the variance in the data. However, given the nature of the task design and very short runs, am I at a significant disadvantage in power when analyzing the data in FSL? For example, would the fact that I only have 8 trials per condition per run result in very underpowered level 1 statistics? In this scenario is the approach of AFNI's concatenation of runs a potential solution, or would it still be recommended that I used the mixed-effects approach of FSL? Relatedly, if you have advice on how to conduct power analyses or design efficiency within FSL - I have not been able to find directions on this topic and would greatly appreciate any info.
Thanks!
Michelle
|