Hello,
We have a subject response contingent analysis, which means that we end up with different numbers of trials in different conditions. Subjects can freely respond A or B (speech production). We are interested in brain regions showing B > A. The issue is that there are different numbers of A and B responses in different conditions (expected given the nature of the task). Some conditions may have as few as 5 B responses, others as many as 20. Under these conditions, is it meaningful to interpret a lack of B>A effect in conditions with low # of B responses and a significant B>A effect in conditions with higher # of B responses? Is there a range for # of trials that is thought to yield reliable parameter estimates?
Thanks in advance,
Malathi
|