 |
 |
I have addressed a similar issue in creating my design scripts. What I do is up front create all possible regressors, eg: Correct Cond A / Incorrect Condition A / Correct Condition B / Incorrect Condition B ... Then I cycle through my behavioral results file (trial specific responses collected at the time of the scanning) and fill in the different regressors. Then I go back and check to see if any regressors are empty. If so, the program removes them and all that is left are the non-empty regressors which go into my model. The next thing to consider is how do you compare correct vs incorrect if the proportions are strongly in favor of correct. So if you have 25 correct trials and 5 incorrect trials. For that I do not have a good answer and am looking forward to what others have to say. I hope this helps, Jason On Mon, Dec 8, 2008 at 11:03 AM, Jason Steffener <[log in to unmask]>wrote: > I have addressed a similar issue in creating my design scripts. What I do > is up front create all possible regressors, eg: Correct Cond A / Incorrect > Condition A / Correct Condition B / Incorrect Condition B ... > Then I cycle through my behavioral results file (trial specific responses > collected at the time of the scanning) and fill in the different regressors. > Then I go back and check to see if any regressors are empty. If so, the > program removes them and all that is left are the non-empty regressors which > go into my model. > > The next thing to consider is how do you compare correct vs incorrect if > the proportions are strongly in favor of correct. So if you have 25 correct > trials and 5 incorrect trials. > For that I do not have a good answer and am looking forward to what others > have to say. > I hope this helps, > Jason > > > On Mon, Dec 8, 2008 at 10:46 AM, Marko Wilke < > [log in to unmask]> wrote: > >> Hi, >> >> there was a response to a similar question a while ago suggesting that you >> include a "dummy mistake" at the very last scan. The effect of this will >> only be taken into account (due to the shift of the HRF) when you are >> already done scanning, so it will not really make a difference (but it may >> allow you to specify your model). Perhaps this is worth a try. >> >> Best, >> Marko >> >> Tully, Joseph (NIH/NIMH) [V] schrieb: >> >>> Hi everyone, >>> >>> >>> My problem is that I'm trying to differentiate between correct & >>> incorrect responses on an fMRI task over the course of 4 runs. The problem >>> is that some people did really well or really poorly in certain runs. For >>> example, subject A may have gotten 100% correct/0% incorrect on the task >>> during run 1. As far as I can tell, this means that it is impossible to >>> Specify 1^st Level and run the program, since the program assumes I've made >>> a mistake and left something in blank rather than the subject literally >>> having 0% incorrect. Is there a way to collapse all 4 runs together so I can >>> still make this comparison between correct & incorrect responses? >>> >>> >>> Thanks, >>> >>> Joe >>> >>> >>> >>> >> >> -- >> ===================================================================== >> Marko Wilke (Dr.med./M.D.) >> [log in to unmask] >> >> Universitäts-Kinderklinik University Children's Hospital >> Abt. III (Neuropädiatrie) Dept. III (Pediatric neurology) >> Hoppe-Seyler-Str. 1, D - 72076 Tübingen >> Tel.: (+49) 07071 29-83416 Fax: (+49) 07071 29-5473 >> ===================================================================== >> > >