Print

Print
Hi,

sorry if I wasn't clear. Well, with Higher-level analysis the only option is Stats+Post-stats. Using First-level with Post-stats only on the previous output doesn't work (Error message: Problem with processing the model: Can't open dummy for reading, child process exited abnormally). It shows the same error no matter if I select the *.gfeat or cope1.feat folder as input. Based on that I assumed that it is not that straightforward and I have to re-run the stats as well. But maybe I am just missing something very simple :-)

I always run FLAME 1 to briefly check the results, but we play with some weak effects, and some of them only appear (or rather survive thresholding) after running FLAME 1+2. We have two groups (n=11 and n=13), with most of the probands having two runs under two conditions, altogether 92 input files. That's why I thought I could run the stats just once and then write a bash script to pre-threshold them with different masks (e.g. with the white matter masked-out) and at different z-threshold.

Within the script I use cluster command and feed it with the output of the smoothest command. However, so far I couldn't find any resource on how to use it properly (since I can feed it either with zstat image or res4d).

I also consider giving up on cluster thresholding and trying TFCE.

Thanks for any advice or pointing out any errors in my thinking :-)

Kind regards

Pavel

On 30 July 2012 09:51, Stephen Smith <[log in to unmask]> wrote:
Hi - I don't quite understand the question - once you have FLAME higher-level zstats then the thresholding via the default FEAT route is straightforward?

If you want to speed up the higher-level stats then you can choose "FLAME 1" or even OLS instead of full FLAME 1+2.

Cheers.



On 28 Jul 2012, at 15:18, Pavel Hok wrote:

Hello FSL Team!

I would like to pre-threshold FLAME zstats without running the whole Higher-level analysis, as it takes a lot of time. I realized that cluster command requires DLH as input, but I am not sure how to use the smoothest command properly. First, I tried using the zstat as input for smoothest, but the results differ from the FEAT output when setting the pre-threshold mask in Post-stats. I found some threads in the mailing list which were dealing with similar issues, for instance calculating DLH from res4d instead of zstat was encouraged. However, I couldn't find any answer to following questions:

1) How is the DOF value for the smoothest command calculated? Should I just use the same value (e.g. 88) as the Feat script (from the log file)?

2) I noticed, that even when the pre-thresholding mask is set, the smoothest masks the res4d file only with whole-brain mask (i.e. mask.nii.gz from stats folder), ignoring the custom mask. So the question is, which mask should I use when running smoothest (whole brain mask or the custom one)?

Thanks!

Pavel



---------------------------------------------------------------------------
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director,  Oxford University FMRIB Centre

FMRIB, JR Hospital, Headington, Oxford  OX3 9DU, UK
+44 (0) 1865 222726  (fax 222717)
[log in to unmask]    http://www.fmrib.ox.ac.uk/~steve
---------------------------------------------------------------------------