fmripower

*Now works with SPM output*

home     email     download     faq     paper     instructions

Introduction


Fmripower was introduced in the 2007 OHBM Poster. New versions will incorporate all of the techniques described in the paper Power Calculations for Group fMRI, which is in NeuroImage (2008 Jan 1;39(1):261-8).



Highlights
  • Matlab based application (requires SPM5 or greater).
  • Uses previous anlyses carried out in FSL or SPM.
  • Power analysis carried out in an ROI approach. Interpretation of power is for an average voxel of that ROI.
  • Reports mean in standard deviation units, which is easy to transfer to a grant application.
  • *New* works with SPM output.
  • *New* automatic design matrix generation for 1-sample, 2-sample and paired t tests. Allows you to put in a single sample size or range of sample sizes for which you would like to calculate power.
  • *New* When user inputs a range of sample sizes the software now automatically generates a power curve for each ROI, which is displayed in the output window.

Who can use fmripower?

The GUI should work on all computer platforms, but it currently has only been tested on Linux/Unix and Macs. Matlab and SPM5 (or higher) are required. FSL users working in Windows will need cygwn, but this is a requirement for FSL, so if you're using FSL you already have it.

How can I get fmripower?

You can download the latest version.

How do I use fmripower?

First download and unzip the fmripower.tgz file and then follow the instructions found here.

How can I get help with fmripower?

If you find any bugs, need help, or have suggestions for future versions email jeanette.mumford@gmail.com.

Can fmripower be used for posthoc power analysis?

Nope. Sorry, power anlayses are only valid for calculating power for future data analyses. See The Abuse of Power: The Pervasive Fallacy of Power Calculations for Data Analysis John M. Hoenig, Dennis M. Heisey The American Statistician, Vol. 55, No. 1 (Feb., 2001), pp. 19-24 for some interesting examples why posthoc power doesn't make sense. If you are interested in evaluating a study that has already been done, the best thing to do is look at the percent change threshold. Tom Nichols has some documentation on percent change threshold (PCT).

email © 2011 Jeanette Mumford