For the power analysis, you need four things: 1. Effect Size 2. Number of Subjects 3. Target False Positive Rate (alpha) 4. Target False Negative Rate (beta)
Given any 3, you can compute the 4th. You can get the Effect Size from the output of the QDEC analysis. Each analysis creates a GLM directory, and there is a directory for each contrast in the GLM dir. In the contrast dir, you will find several files, but the important ones for this are the gamma.mgh and the gammavar.mgh. The gammavar is the square of the error bar (ie, t=gamma/sqrt(gammavar)). The gamma will not change as you add subjects (at least in expectation). The gammavar will drop linearly with the number of subjects (again in expectation). The effect size will then be gamma/sqrt(gammavar*Npilot), where Npilot is the number subjects in the pilot study. In a new study with Nnew subjects, the expected t will be tnew = gamma/sqrt(gammavar*Npilot/Nnew)
doug
Emily Rogalski wrote:
How would you complete a power analysis for a whole brain group comparison qdec analysis?
Thanks, Emily
-- Emily J Rogalski, PhD Assistant Research Professor Northwestern University Cognitive Neurology and Alzheimer's Disease Center (CNADC) 320 E Superior Street Searle Building 11th Floor Chicago, IL 60611 312-503-1155 phone 312-908-8789 fax erogalski@gmail.com mailto:erogalski@gmail.com
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer