I recommend that you read the recent email thread (on this list) about power analyses to see why what you are proposing is not sound. See the links here: https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind1508&L=SPM&F=&S=&P=669027
________________________________________
From: SPM (Statistical Parametric Mapping) [[log in to unmask]] on behalf of Natalia Yakunina [[log in to unmask]]
Sent: Thursday, September 3, 2015 03:17
To: [log in to unmask]
Subject: [SPM] Improving significance
Dear experts,
I performed several multiple regression analyses in SPM for several stimulation protocols on the same group of subject, and in spite of a relatively large number of subjects (37), nothing or very little survives FWE threshold (which is pretty high T>7.5). FDR thresholding doesn't work (gives infinity), and uncorrected results cannot be reported. Is there a way to increase the significance of the results? We intentionally recruited more subject in order to get more significance in the results, but apparently it did not matter.
Thank you beforehand.
Sincerely,
Natalia
|