<div dir="ltr">Thanks for your reply.<br><br>My data is functional localizer of the FFA (blocks of faces vs. blocks of objects). To avoid contamination I have 15 sec fixation between blocks. The task is 1-back. So, nothing special (more or less like your experiment :)<br>
<br>1. In preprocessing I run slice time correction so it should be a problem.<br>3. Experiments session start is triggered by the scanner and the session ends exactly with scan end.<br><br>Your proposal of ER analysis is interesting. I will do it. Actually, I can run a stadard GLM for each TR of my block as though it was event related design and then make t-cotntrasts for each event. However, for first TRs I may get better results because they are not contaminated by previous signal (remember, I have a fixation beforehand). Maybe this is also the reason for my better prediction issue? BTW, what happens when I analayze a block desing data as ER. I have no optseq so model fit should decrease. Correct?<br>
<br>Vadim<br><br><br><div class="gmail_quote">On Sun, May 24, 2009 at 6:03 AM, Yaroslav Halchenko <span dir="ltr"><<a href="mailto:debian@onerussian.com">debian@onerussian.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
Well -- it is indeed an interesting effect ;) various factors could<br>
contribute to it and it is hard to comment without specifics of the<br>
experiment/task/acquisition<br>
<br>
1. volume is not an instantaneous measure, so it might be that the slice<br>
of interest is actually delayed already by a TR*(1-1/number_of_slices)<br>
<br>
2. who knows -- pre-dip effect exists and some people claim that it has<br>
better spatial locality than full-blown response<br>
<br>
3. if trial is not magnet-pulse triggered, it is easy to slowly<br>
accumulate some lag by the end of the series<br>
<br>
etc<br>
<br>
the best way to see what is happening imho is to do ER-like analysis by<br>
taking multiple volumes after each onset and then looking at the<br>
sensitivities for interesting voxels across time; and also to look at<br>
per-stimuli plots --- where is the most difference between responses to<br>
two conditions. Noone should fully believe into a static homogeneous HRF<br>
across all voxels/conditions.<br>
<br>
Michael, should we put your face/shoe pictures online -- they might<br>
benefit some people ;)<br>
<div class="im"><br>
On Sat, 23 May 2009, Vadim Axel wrote:<br>
<br>
> Hi,<br>
> I am getting consistently better prediction when I don't make<br>
> haemodynamic lag shift. Have you ever experienced something similar? I<br>
> would expect the opposite results. I have closely examined the code and<br>
> there are no bugs there.<br>
> Thanks,<br>
> Vadim<br>
</div><font color="#888888">--<br>
.-.<br>
=------------------------------ /v\ ----------------------------=<br>
Keep in touch // \\ (yoh@|www.)<a href="http://onerussian.com" target="_blank">onerussian.com</a><br>
Yaroslav Halchenko /( )\ ICQ#: 60653192<br>
Linux User ^^-^^ [175555]<br>
</font><div><div></div><div class="h5"><br>
<br>
<br>
_______________________________________________<br>
Pkg-ExpPsy-PyMVPA mailing list<br>
<a href="mailto:Pkg-ExpPsy-PyMVPA@lists.alioth.debian.org">Pkg-ExpPsy-PyMVPA@lists.alioth.debian.org</a><br>
<a href="http://lists.alioth.debian.org/mailman/listinfo/pkg-exppsy-pymvpa" target="_blank">http://lists.alioth.debian.org/mailman/listinfo/pkg-exppsy-pymvpa</a><br>
</div></div></blockquote></div><br></div>