My lab has had the same problem when identifying trials with errors, and --
if memory serves me -- I think we dealt with the issue as follows.
For subjects that had no errors, we specified that an error occurred during
the last acquired volume. This has no practical significance in modeling
effects, as the hemodynamic delay precludes such a fictional event from
affecting the modeled activation, but it does (I think) allow the condition
to be included in the design matrix. Since we were interested in group
activation from the error trials, we only carry the contrasts or parameter
estimates forward to the second level (random effects) analysis from
subjects who actually did have error trials.
It's been awhile since we dealt with this and I'm not at all sure that my
memory is accurate, so perhaps someone can try this procedure and post to
the list whether it works.
Doug
At 02:19 PM 3/12/2007 +0100, Volkmar Glauche wrote:
>Dear Ninad,
>
>a late follow-up on your question (I've just come across the same problem)
>- but without a solution yet:
>1) If an event does not occur at all, then SPM will not create any column
>for it in your design matrix. There is no way around it - this column
>would be all-zero anyway.
>2) If you have a parametric modulation, which does not modulate anything
>(e.g. a constant parameter for all occurences of your event), then only
>the main effect regressor will be built, since your parametric modulation
>would be an all-zero column.
>Currently, there is no easy way to automatically specify contrasts when
>conditions or modulations are missing. A way out of this would be to build
>contrasts based on column names/condition names instead of column numbers,
>but this will require a major redesign of SPMs contrast manager batch.
>
>Volkmar
>
>On Wed, 3 Jan 2007, Ninad Gujar wrote:
>
>>Hi,
>>
>>Happy New Year to everyone !
>>
>>I am still trying to solve the same problem from last year.
>>
>>Our study consists of five sessions, each with 136 volumes. Now, each
>>session had 4 events.
>>
>>Whenever there are no events to model e.g. no misses I tried embedding in
>>the script the parameter to accept the value 'none' and SPM does accept the
>>value as 'none' . Is this the right way to do it ?
>>
>>Now, some of the session events have only one value to model. i.e. only
>>onemiss
>>SPM does not create the column for the regressor and only include one column
>>for the event.
>>
>>I checked in the SPM structure and all the values including the onset and
>>the corresponding regressor seem to exist.
>>Is there a problem if my event has only one value ?
>>
>>Some of our code: (for one event it looks like this)
>>
>> SPM.Sess(ses).U(count).P(1).name = 'ArousalScore';
>> SPM.Sess(ses).U(count).P(1).h = 1; % order of polynomial
>>expansion
>> SPM.Sess(ses).U(count).P(1).i = [1 2];
>> SPM.Sess (ses).U(count).P(1).P = [6.65];
>> SPM.Sess(ses).U(count).ons = [274];
>> SPM.Sess(ses).U(count).dur = 0;
>>
>>I need to create the column for this since while creating the contrasts at a
>>much broader level e.g. modeling just the misses I combine all the misses
>>from all sessions.
>>
>>Any help would be most appreciated !
>>
>>Thanks,
>>
>>Ninad Gujar
>>
>
>--
>Volkmar Glauche
>-
>Department of Neurology [log in to unmask]
>Universitaetsklinikum Freiburg Phone 49(0)761-270-5331
>Breisacher Str. 64 Fax 49(0)761-270-5416
>79106 Freiburg http://fbi.uniklinik-freiburg.de/
Dept. of Communication Sciences & Disorders
Northwestern University
2240 Campus Drive
Frances Searle Building, Room 2-356
Evanston, IL 60208
phone 847-467-1549
fax 847-491-4975
email: [log in to unmask]
|
