Dear Yiwen,
On Fri, Nov 27, 2009 at 1:51 PM, Yiwen Li <[log in to unmask]> wrote:
> Dear all,
>
> My experiment is a bilateral tactile discrimination task measured with the
> 275-channel CTF system. After a long struggle failing to localize bilateral
> primary somatosensory sources using different beamforming techniques, I have
> to say that I am now very happy with the results SPM8's MSP returned. The
> sources are very similar to the fMRI results I got from the same
> experimental design with the same group of subjects. Many thanks to the
> program developers!
You are welcome :-) Beamforming is not a good technique for
reconstructing evoked responses. It's more suitable for oscillatory
activity.
>
> But now I have a problem with setting an optimal time window for group
> inversion. My original trial length is 5.5s (-1.5 to 4s, -1s is visual
> instruction onset, 0s to 0.6s is tactile stimulation, 1.6s is instruction
> for button presses). My first approach was cutting the trials [0 0.3]s and
> group inversion gave me very clear bilateral SI and SII sources. I am just
> wondering whether I should feed the group inversion a longer data piece as
> was suggested in the manual, e.g. to include some baseline? As different
> brain sources are involved along the trial, I am not sure how this will
> influence the results.
>
If you look at the web version of the manual
(http://www.fil.ion.ucl.ac.uk/spm/doc/spm8_manual.pdf) on page 119 it
now says something slightly different which is based on recently
accumulated experience. Your choice of time window sounds perfectly
sensible to me and if it produces the results you expect there is no
reason to change it. Also there is no reason to include the baseline
if you don't need baseline images for your stats.
> The second problem is, I couldn't run a 2nd-level t-test. I specified a 2x2
> factorial design using the 8 mm smoothed images. I defined the f contrasts
> as the following:
> main effect: eye(4)
> first factor: [1 1 -1 -1]
> second factor: [1 -1 1 -1]
>
eye(4) doesn't sound like a good idea as all your images are strictly
positive. This has been discussed before
(https://www.jiscmail.ac.uk/cgi-bin/wa.exe?A2=ind0909&L=SPM&P=R11297).
The other two contrasts are OK.
> I had no problem running all the f tests but I run into errors when I define
> the t contrast:
>
> In spm_invTcdf at 93
> In spm_u at 31
> In spm_uc_RF at 30
> In spm_uc at 39
> In spm_getSPM at 705
> In spm_results_ui at 277
> ??? Error using ==> betainc at 42
> X must be in the interval [0,1]. etc...
>
> I saw the same error reported in previous emails due to using unsmoothed
> images. I thought that if I get some results from the f contrast testing
> each factor, which is equivalent to a two-sided t-test, I should also be
> able to get some results from a one-sided t-test?
>
Here I'm not sure what the problem is. Perhaps Guillaume will have a
better idea.
Best,
Vladimir
> Thank you very much for your help in advance!
>
>
>
> Best regards,
> Yiwen
>
|