search

cvnlab / GLMsingle

A toolbox for accurate single-trial estimates in fMRI time-series data
BSD 3-Clause "New" or "Revised" License
97 stars 42 forks source link

if user turns off certain routines within GLMsingle (e.g. GLMdenoise), output beta filenames will be misleading #22

Closed jacob-prince closed 2 years ago

jacob-prince commented 2 years ago

currently, in both matlab and python implementations, output filenames are hard-coded as follows:

thus, if users output TYPEC betas but turn off HRF fitting, or, output TYPED betas but turn off either HRF fitting or GLMdenoise, the output filenames will be misleading.

a bug fix here would modify the output filenames appropriately to reflect the features that the user has turned on/off. so if they ran only ridge regression with a canonical HRF and no denoising, the output would be "TYPED_RR" or even "TYPED_ASSUMEHRF_RR" if we wanted to be extra descriptive. similar functionality could even be added for the LSS cases...

it's possible that users should be left to "fend for themselves" here, but having files lying around that are mislabeled could create serious confusion if users don't create descriptive documentation for themselves to make sure they remember which output files correspond to a specific set of input parameters.

here are the lines with the hard-coded output filenames:

image

image

kendrickkay commented 2 years ago

Hmm... Yeah, I agree this is a tricky issue. I'm not sure what's least bad? Make generic filenames? Make the user specify the filenames themselves? Make extra-informative custom filenames? These all have downsides.... ?

kendrickkay commented 2 years ago

Hmm, one idea is that we could make the filenames generic and then have the code save a little text file that documents exactly what was run/options/etc. to generate that version of the results.

iancharest commented 2 years ago

I think this is a case for users knowing what they do. Overall everything is carefully documented so should not be a major issue.

kendrickkay commented 2 years ago

Ian and I discussed and we feel it's okay to let users fend for themselves (i.e. keep track of what they are doing).

We feel the alternative (i.e. tailoring filenames) is even a worse solution.