Hello,
I have a question about statistical power in an event-related experiment. In the experiment we blocked trials of a particular condition together, but within that block the trials were presented in an event-related manner (jittered with baseline trials).
In our experiment there were 8 conditions. We split the trials for each condition into 2 10-trials blocks so that we could randomize the conditions better across the 4 experimental runs. For example,
Run1: cond1a, cond8b, cond4b, cond6a
Run2: cond7a, cond2b, cond3a, cond5b
Run3: cond8a, cond5a, cond7b, cond6b
Run4: cond2a, cond1b, cond3b, cond4a
My question is: Assuming that the number of useable trials (e.g., 13/20 trials) was equal across conditions, would it matter if in some conditions the useable trials were more spaced out and in other conditions the useable trials were more clumped together?
Would my statistical power be higher when the trials are more clumped together that when they are more spaced out?
A reviewer has suggested that because 1) I blocked trials into the 10-trial blocks (and did not do a completely randomized design) and 2) in some conditions trials were more spaced out than in other conditions, that the clumped together conditions have more statistical power than the spaced out conditions.
This doesn’t make sense to me because I thought that spaced out might actually let you model the HRF better than clumped together.
Any advice or references to what the reviewer might be talking about are appreciated.
Thank you,
Christine
