 |
 |
Dear Daniel and Rik, Thank you for your replay. I was thinking in the begging to use FIR because it could have been more sensitive to the response shape, but in the end we decided to use the HRF and its derivatives. We choose the HRF and its derivatives, because we were assuming that it might be a difference in the response shape (time or dispersion) between groups, which by using the canonical HRF alone, we could not capture. I have a jittered ITI (inter-trial interval) of 0.5 to 1.5s and each trial was present on the screen for 2.5s (TR was 2.3s). In my model (1st level), each trials was modeled as events (dur=0). I suppose that I can use the derivatives with this setup, as my SOA (I believe is the time between two trials, is 3-4s), or I'm wrong? I'm not sure: how exactly can I check for the HRF shape efficiency? Moreover, I defined a functional mask for one region of interest and extracted the data using Marsbar, than I plotted the sum of the 3 basis function per groups. I saw that one of the groups showed a later and lower BOLD response that the other group. My From question comes from this point, as I wanted to test if this difference is significant. I was thinking that doing an F-test, as I specified earlier, it should tell me if these two groups differ and where. Is that true or is there a different way to test for this? Kind regards, Ramona On Fri, Mar 27, 2009 at 5:17 PM, Rik Henson <[log in to unmask]>wrote: > Romona - > > Your F-contrast is correct for testing any difference in response shapes > between the two groups, where that shape can be captured by your three basis > functions. Daniel's suggestion of an FIR will capture an even greater range > of shape differences, though possibly at reduced power if those differences > are within the subspace defined by the canonical and its two partial > derivatives (though the nice thing about an FIR at the 2nd-level is that you > can still create more focussed, ie potentially more sensitive, F-contrasts > akin to the canonical and its derivatives - see Chapter 30 of SPM5 manual, > page 248 onwards). > > HOWEVER, tests of shape differences (vs baseline) require that you can > estimate the HRF shape efficiently, which means that you must have a > jittered SOA (eg of order 1-12s), eg through appreciable ration of null > events. There is no point trying to estimate response shapes (vs baseline) > if you have a short SOA design. > > Rik > > > Dear Ramona, >> >> Have you considered using a finite impulse response (FIR) model to analyze >> your event-related data? Such a model would allow you to estimate the >> average stimulus-locked BOLD response to each of your trial types across >> time. Comparisons of these response shapes for your 2 subject groups might >> then be more easily (i.e., visually) interpretable. >> >> Hope this helps, >> Daniel >> >> Daniel Weissman, Ph.D. >> Assistant Professor >> Department of Psychology >> University of Michigan >> Room 4052 >> 1012 East Hall >> 530 Church Street >> Ann Arbor, MI 48109 >> ________________________________________ >> From: SPM (Statistical Parametric Mapping) [[log in to unmask]] On Behalf >> Of Liliana Demenescu [[log in to unmask]] >> Sent: Friday, March 27, 2009 11:22 AM >> To: [log in to unmask] >> Subject: [SPM] Hemodynamic response shape testing. >> >> Dear all, >> >> I have an event-related design. The 1st level included the canonical HRF >> and its time and dispersion derivatives, because we want to look if there >> are difference in response shape between 2 groups. At the 1st level >> analysis, I defined a t-contrast for each basis function (HRF, TD and DD). >> The 2nd level analysis employed an ANOVA 3 (basis functions) x 2 (groups). I >> did an F test defined as: >> 1 -1 0 0 0 0 >> 0 0 1 -1 0 0 >> 0 0 0 0 1 -1 >> (group1 - group2: HRF; group1 - group2: time der. and group1 - group2: >> dispersion der.). This tells me which voxels have different activity between >> the 2 groups with respect to either amplitude, either time, or dispersion. I >> plotted the results, but I found it difficult to understand/explain, as it's >> an F test. Moreover, as I understood from Friston's paper (Friston et al. >> 1998, Neuroimage, 7:30-40) this tells me how well the model fits or have I >> misunderstood the things? >> >> My question is how can I look for differences in response shapes between 2 >> groups combining the 3 basis functions? >> >> Any suggestion is more than welcome. >> >> Thank you! >> >> Kind regards, >> Ramona >> >> > > > -- > > ------------------------------------------------------- > > DR RICHARD HENSON > MRC Cognition & Brain Sciences Unit > 15 Chaucer Road > Cambridge, CB2 7EF > England > EMAIL: [log in to unmask] > URL: http://www.mrc-cbu.cam.ac.uk/people/rik.henson/personal > > TEL +44 (0)1223 355 294 x522 > FAX +44 (0)1223 359 062 > MOB +44 (0)794 1377 345 > > ------------------------------------------------------- >