 |
 |
Hi Ari, I see your problem and there likely is no optimal solution, so you might end up having to choose from two lesser evils. I had probably not dared to suggest using one dummy regressor for the last scan if the idea had not been Karl's in the first place ;) Due to the hemodynamic delay, you should not really see an effect in that regressor; it will, however, prevent the power issue I mentioned. Further, you can also not use that regressor if you do not specify it, so this problem will be similar if you choose the alternative you originally suggested. Perhaps you can specify an additional regressor that models some kind of baseline in all subjects, regardless of them making errors or not, and use that as a reference? Cheers, Marko Ari Pinar wrote: > Hi Marko, > > > Thanks for your response. > > The concern I had with option 2 is focused on the interpretation of the > contrast. > > Say I want to contrast two types of errors, errorA and errorB. > > For one subject, there are no events corresponding to errorB in one session. > > If I model errorB with a regressor in which the only event is the last > time point, then the contrast will be errorA vs last [timepoint], which > doesn't make much sense to me. > > Am I missing something here? > > > Thanks again, > > > Ari > > > Ari Pinar > > PhD Candidate > Monash Clinical & Imaging Neuroscience > School of Psychology and Psychiatry & > Monash Biomedical Imaging > Monash University > > A: 770 Blackburn Rd, Clayton, 3168, Vic, Australia > E: [log in to unmask] <mailto:[log in to unmask]> > > > On 17 October 2013 23:48, Marko Wilke <[log in to unmask] > <mailto:[log in to unmask]>> wrote: > > Ari, > > I would opt for option 2, for the simple fact that you "win" degrees > of freedom in those that do not make errors, which then again may > bias your analyses later-on. Not by much (see PMID 22036679 for some > pretty extensive simulations), but it may still unbalance your > design nonetheless. > > Cheers, > Marko > > > Ari Pinar wrote: > > Hi all, > > I have a fMRI task comprising 4 conditions and 6 sessions. > > I am trying to build a first-level design matrix for each > subject in which distinct regressors encode correct responses > and errors in the different conditions. > > The problem is that for some subjects and some sessions, no > errors are committed. This means that my error regressors are > empty for some sessions (e.g., there may be events for sessions > 1, 3, 4 and 5 but not 2 and 6). > > It seems like there are two ways of dealing with this: > 1 – don’t specify a regressor if the error condition contains no > events in a given session > 2 – create a dummy regressor in which there is a single value > for the last timepoint (as suggested on another post) > > I have two questions: > 1 – is one approach better than the other? Seems to me like > option 1 is preferable. > 2 – what is the best way of handling contrasts in this case? I > could specify separate contrast estimates for each session, then > create an average contrast across only the sessions in which the > contrast is valid. However, if there are data for 3 sessions in > one subject and data for 6 sessions in another subject, will > this create scaling problems or does SPM8 already account for this? > > Thanks for your help, > > Ari > > > -- > ______________________________________________________ > PD Dr. med. Marko Wilke > Facharzt für Kinder- und Jugendmedizin > Leiter, Experimentelle Pädiatrische Neurobildgebung > Universitäts-Kinderklinik > Abt. III (Neuropädiatrie) > > > Marko Wilke, MD, PhD > Pediatrician > Head, Experimental Pediatric Neuroimaging > University Children's Hospital > Dept. III (Pediatric Neurology) > > > Hoppe-Seyler-Str. 1 > D - 72076 Tübingen, Germany > Tel. +49 7071 29-83416 <tel:%2B49%207071%2029-83416> > Fax +49 7071 29-5473 <tel:%2B49%207071%2029-5473> > [log in to unmask] > <mailto:[log in to unmask]> > > http://www.medizin.uni-__tuebingen.de/kinder/epn/ > <http://www.medizin.uni-tuebingen.de/kinder/epn/> > ______________________________________________________ > > -- ____________________________________________________ PD Dr. med. Marko Wilke Facharzt für Kinder- und Jugendmedizin Leiter, Experimentelle Pädiatrische Neurobildgebung Universitäts-Kinderklinik Abt. III (Neuropädiatrie) Marko Wilke, MD, PhD Pediatrician Head, Experimental Pediatric Neuroimaging University Children's Hospital Dept. III (Pediatric Neurology) Hoppe-Seyler-Str. 1 D - 72076 Tübingen, Germany Tel. +49 7071 29-83416 Fax +49 7071 29-5473 [log in to unmask] http://www.medizin.uni-tuebingen.de/kinder/epn/ ____________________________________________________