Dear all.
I have a stats question regarding flexible factorials with uneven
sample sizes. I have 15 subjects, each with hits and false-alarm
trials/images from my decision-making task. I am performing a flexible
factorial using subject as one factor, and task condition as the
other. I am interested in seeing how regions differ between hit trials
and false-alarm trials.
However, each subject has a different proportion of hits and
false-alarms. Is it still statistically-sound to perform such a
contrast when the number of hits and false-alarm trial/images differ
across the 15 subjects? Or, must as adjustment to made such as
weightings for each subject?
Muhsin.
|