I still have the question whether a value of '0' in a parametric
modulation means that the corresponding onset will not be taken in
account for the pmod regressor?
And is it correct to collapse all conditions together and use
parametric modulation to detect gradual changes between them in a
Your help is much appreciated.
2009/1/28 Dorian P <[log in to unmask]>:
> Dear Volkmar,
> Your idea made me think of adding more and more pametric modulations
> for a single condition. But would it be possible to skip a level of
> modulation, i.e.
> pmod(1) = 1-2-3-4-5
> pmod(2) = 1-2-3-4(-0) (!!?)
> Should I put '0' to a level to account the remaining 4 levels or
> should I add another condition with a new parametric modulation with
> only 4 levels of modulation?
> And one more question. Is it reasonable to use modulation 2-1-1-2 with
> contrast "-1" to account for the modulation 1-2-2-1? I am asking
> because they are not linear as the previous example.
> Thank you very much for you help.
> 2009/1/28 Volkmar Glauche <[log in to unmask]>:
>> Dear Dorian,
>> see inline comments...
>> Am Mittwoch, den 28.01.2009, 11:50 +0100 schrieb Dorian P:
>>> Une small addition:
>>> If I need to try modulation 1-2-3-4-5 and see the regressor effect
>>> with contrast [0 1],
>>> can I just run contrast [0 -1] to save another GLM with modulation 5-4-3-2-1 ?
>>> > I have a simple question. I need to run around 5 GLMs for each subject
>>> > (N = 19) with different parametric modulation values for different
>>> > variables.
>>> > My question is: Is there a way to save time by using some calculations
>>> > previously done for the subject?
>>> > I don't know the steps of a GLM, but thinking of SPM, it first
>>> > establishes the number of resels and then runs the analysis, and
>>> > thinking that the volumes to analyze are the same for each subject,
>>> > and though number of resels should be the same... can this be useful
>>> > to the next GLM? Any clue how to save time?
>> No. Even if the number of voxels that are considered is the same, the
>> smoothness of the error (and thus the number of resels) will depend on
>> the model. However, you should model all effects that you expect to be
>> in your data in a single model, not in separate models. There may be
>> reasons to run 2 or 3 models (e.g. some experiments may be modeled
>> either as block or event related), but there is no way to shorten the
>> analysis time.
>> 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/