thanks for the confirmation. This is much clearer now. I had realised my confusion and mistake after browsing the mailing list some more but was mixed up for some reason with the old "SPM-way" of having to activate a contrast (mistaking c for contrast not cluster threshold).
Anyways, thanks again for your help! Much appreciated!
2.3 is the value that will be used as the cluster-forming
threshold for your f-stat(s), if you don't want
cluster-corrected results then you can use the --T2
option or even just -x for voxelwise results. The F-test
is a statistic type, --T2, -F, -x etc are just various
methods for dealing with multiple-comparisons.
I hope this helps,
From: Muller, Sven (NIH/NIMH) [V]
Sent: Thursday, April 11, 2013 5:59 AM
To: [log in to unmask]
Subject: Re: clarification on F-test
this seems to have worked but you managed to confuse me a little more with regards to the F-test. Sorry. :-) You say ("e.g. -F 2.3"). I thought that if I only specified one F-test in my design and checked the three boxes, this would be a combined F-test. If what you are writing is correct in my case, then should I write -F 1.2.3 in order to get the full F? Sorry, I don't understand the "2.3"....
Another quick question:
Can I combine the F-test with the --T2 option or are these exclusive? I am not sure I quite understand whether these two options can be run together in one analysis or whether they should be run separately.
Thanks so much for your help,
It looks like you are using a lower case ( -f ) for the
F-test cluster threshold option, this should be a
capital F ( e.g. -F 2.3 ) - can you try this and let us
know if the error changes?
> Dear FSLers,
> I am new to FSL and am currently analysing 4 groups DTI TBSS data and have
> worked my way through the tutorial and posts on the mailing list. However,
> I am still confused with the F-Test.
> If I include the -f design.fts option but not specify an -F [number]
> option, I get an error message that no specific option was selected but if
> I do say -F 1 then I get another error message. So concretely, I had used
> the example on the TBSS stats webpage:
> randomise -i all_FA_skeletonised -o tbss -m mean_FA_skeleton_mask -d
> model_test1.mat -t model_test1.con -f model_test1.fts -n 5000 -c 2 -V
> Loading Data:
> Data loaded
> 1.16979e+44 permutations required for exhaustive test of f-test 1
> Doing 5000 random permutations
> Warning! No output options selected. Outputing raw tstat only
> I had then tried
> randomise options: -i all_FA_skeletonised -o tbss -m mean_FA_skeleton_mask
> -d model_test1.mat -t model_test1.con -f model_test1.fts -n 5000 -c 2 -f 1
> but got
> Loading Data: ERROR: Program failed
> An exception has been thrown
> Unable to open 1
> If my design matrix is such that
> 1 -1 0 0 check
> 1 0 -1 0 check
> 1 0 0 -1 check
> …… set of remaining t-contrasts here without selection of F, e.g.,
> 0 0 1 -1 not checked
> … etc
> What option should I use in randomize to get the output for my F-test?
> Should I use a combination of these to specify the contrast or can I
> ignore the error and check this once the statistics have run?
> Many thanks in advance for any suggestions!