Hi everyone,

I have a question regarding 3dundump that's been nagging at me for awhile, and I couldn't seem to find the answer for myself.

Specifically, when providing an input file to 3dUndump (in my case various voxel-wise parameters from my analysis), some of the values are NaN (from MatLab). How does 3dundump handle 'NaN' as an input value? Does it treat this as no value being specified for the voxel, defaulting to 'dval'? Or does it just exclude that particular voxel value from any subsequent calculations (such as 3dANOVA, ttest, etc)?

The other issue is that I have multiple tlrc'ed datasets, and some contain voxels (with undumped values) in tlrc space that others do not. When I go about doing ANOVA, I suppose I should restrict the ANOVA dataset to only those voxels contained in all the datasets. Otherwise, you have some missing voxels assigned a value of say 0 through 'fval'.

However it seems the best thing would be to specify 'NaN' as a value in order to omit that particular value from any statistical analysis, as opposed to having to assign a value of say 0 with -dval or -fval which would mess up the statistics. I tried using NaN as a value for -fval and -dval once and I remember it not working, so I'm just wondering if anything like this is possible.

I hope that made some sense.. if not, let me know.

Thanks and Happy New Year!

John