 |
 |
Nils, could it be that if you put both EVs in one single regression model then the two are eating off each other's variance? If you put them in two separate models, each will report all the variance associated with each EV. Once you put them together, then the regression coefficients will only report, for each EV, what is unique to it. If there is enough overlap (i.e. common variance) then you get exactly the scenario you are describing. just a possibility.. cheers martin Nils Richter wrote: > Hi, > I'm sorry, but I don't think I expressed myself clearly. > The timecourses were not averaged for the one EV models. There were two one EV models, that were run seperately. Upon visual comparison they also looked different. > When I put the two timecourses into one model as EV1=timecourse of stimulus and EV2= timecourse of response the contrasts for each of them showed almost no activation (as opposed to sensible activation when running them seperately) and the timecourse comparing them (contrast 3: EV1=1, EV2=-1) showed weird subcortical activation. > Now the latter doesn't really surprise me, maybe there are no significant differences between the timecourses at the thresholds I used, but I would have assumed that activation pattern for the contrasts running just the individual timecourses would look like the activation for timecourses each run in their own model. > I have the feeling that I am missing something really obvious here regarding the statistics involved in using multiple contrasts. > Cheers, > Nils > > > -------- Original-Nachricht -------- > >> Datum: Thu, 6 Nov 2008 08:12:48 +0000 >> Von: Steve Smith <[log in to unmask]> >> An: [log in to unmask] >> Betreff: [FSL] >> > > >> Hi - I think you're saying that you have two somewhat similar >> timecourses; when you in some way averaged them into a single EV you >> found reasonable activation, but when you separated them, they didn't >> show separate activations (or anything under the differential >> contrast). This isn't necessarily surprising if they are similar to >> each other - assuming they are both accurate models, then you can get >> back to the 'average' result with a [1 1] contrast - but it they are >> very similar to each other, all the other contrasts will be very low >> efficiency (see the efficieny reporting in the FEAT model GUI, and >> also the Smith NeuroImage paper on efficiency). >> >> Cheers. >> >> >> On 5 Nov 2008, at 11:42, Nils Richter wrote: >> >> >>> <[log in to unmask] >>> >>> <[log in to unmask]> >>> <[log in to unmask]> >>> Subject: comparing timecourses >>> To: "FSL - FMRIB's Software Library" <[log in to unmask]> >>> X-Authenticated: #18468792 >>> X-Flags: 0001 >>> X-Mailer: WWW-Mail 6100 (Global Message Exchange) >>> X-Priority: 3 >>> X-Provags-ID: V01U2FsdGVkX18ZsK4mvOIZYFi5iUYnGlfDMbo9hx+dHqYFZR9eEG >>> 0qgVKVapS8eZxUIqFHYbAVJd7E4UOvWAUpgw== >>> Content-Transfer-Encoding: 7bit >>> X-GMX-UID: vbQocbcnYW0tXty8eGdp00V8amthc5tR >>> X-FuHaFi: 0.76 >>> X-CCLRC-SPAM-report: 0 : >>> X-Scanned-By: MIMEDefang 2.57 on 130.246.193.104 >>> >>> Hi everybody, >>> >>> I have a problem regarding the comparison of two different >>> timecourses. One is the timecourse of stimulus onset and the other >>> is the timecourse of response onset. Because the time from stimulus >>> to response varies from trial to trial the shift between the >>> timecourses ist not the same for each trial. >>> I want to see if they yield different activation. Now I have run >>> both timecourses in one EV models and on visual inspection the >>> activation appears to be somewhat different between the two. >>> To quantify this difference I have tried to run the two timecourses >>> as EVs in the following two EV model: >>> >>> EV1: stimulus timecourse >>> EV2: response timecourse >>> >>> Contrast 1: EV1=1, EV2=0 >>> Contrast 2: EV1=0, EV2=1 >>> Contrast 3: EV1=1, EV2=-1 >>> >>> The three contrasts all show very little activation and the >>> contrasts 1 and 2 bear no resemblance to the contrasts in the one ev >>> models, which I can't make sense of. >>> There appears to be some sort of interaction between the contrasts. >>> Cheers, >>> Nils >>> -- >>> Ist Ihr Browser Vista-kompatibel? Jetzt die neuesten >>> Browser-Versionen downloaden: http://www.gmx.net/de/go/browser >>> >>> >> --------------------------------------------------------------------------- >> 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 >> --------------------------------------------------------------------------- >> > >