Re: [Marsbar-users] how to create and use a 'weighting' ROI
Status: Beta
Brought to you by:
matthewbrett
From: Matthew B. <mat...@gm...> - 2010-09-29 17:05:47
|
Hi, > So as I understand if I ask MARSBAR to compute a simple mean from my roi > it will give me a mean as if my roi was binary and if I ask it to give > me a wtmean then it will give me the weighted mean. So I did and I > checked with two rois identical in size and position but different in > their values for each voxel. One was binary and the other had different > integers instead of the ones. I run the batch and ask for a 'wtmean' > instead of 'mean' as a summary function. Then when I look in the > statistics the "t statistic" is the same... Ah - the 'wtmean' stuff has not been very well tested. For the ROI with 'integers' instead of ones, were they all the same integer (instead of 1s) or lots of different integers? If they were all the same, then the t-statistic will be the same, because you've just scaled the data by a constant, and thence the signal to noise hasn't changed... Best, Matthew |