Thank you both for answering.
Jan, I didn't catch very well the idea. Are you proposing to split the
actual trials in different conditions?
This would be problematic. I have a memory experiment with 5 levels.
In one parametric modulation I want to catch 1-2-3-4-5. In another one
I want to leave out "5" and catch 1-2-3-4 only. I cannot catch the
first pmod if I split level 5 to another condition. Finding a way to
exclude trials can be very useful to try modulations in different
ways, i.e. all response levels, or with only correct responses, etc.
Zeithamova, I didn't understand well. The problem is that if you
repeat a single onset value in two different conditions, the whole GLM
becomes unusable because betas cannot be uniqely identified. And
modulations can set up only after you specify a condition. That's why
I collapsed all values to a single condition and played around with
modulations.
Best regards.
Dorian.
2009/2/5 Jan Gläscher <[log in to unmask]>:
> I was thinking of another condition, but you would also have to remove
> these onsets from the onset vector that now includes all trials. I was
> thinking of it along the lines of a nuisance regressor that captures the
> trials you want to model (e.g. missing response trials), but not in your
> regressors of interest. Does that make sense?
>
> Jan
>
> Dorian P. wrote:
>> Hi Jan,
>>
>> Are you suggestingt another modulation or another condition?
>>
>> I have a collapsed condition including all trials and then using
>> parametric modulation to distinguish between them. If I add another
>> condition it will repeat some of the onset vectors of the first and
>> the betas will not be uniqely specified. At that point I am not sure
>> if parametric modulations would still be available for contrasting (or
>> correctly evaluated).
>>
>> Dorian.
>>
>>
>>
>> 2009/2/5 Jan Gläscher <[log in to unmask]>:
>>> Hi Dorian,
>>>
>>> why not model those trials in a different regressor?
>>>
>>> Cheers,
>>> Jan
>>>
>>> Dorian P. wrote:
>>>> Hi SPM list,
>>>>
>>>> I made this question some time ago but still didn't get a response.
>>>>
>>>> I want to leave out some of the trials in a parametric modulation
>>>> regressor. Is there a way to do this?
>>>>
>>>> I would have thought to put "0" in the modulation of those trials but
>>>> it doesn't seem clever if the other values are "1", "2", "3", "4". It
>>>> may take the zero and use it in the modulation as simply less then
>>>> one.
>>>>
>>>> Any clue please?
>>>>
>>>> Dorian.
>>>> Ruhr University Bochum.
>>>>
>>> --
>>> Jan Gläscher, Ph.D. Div. Humanities & Social Sciences
>>> +1 (626) 395-3898 (office) Caltech, Broad Center, M/C 114-96
>>> +1 (626) 395-2000 (fax) 1200 E. California Blvd
>>> [log in to unmask] Pasadena, CA 91125
>>>
>>>
>>
>
> --
> Jan Gläscher, Ph.D. Div. Humanities & Social Sciences
> +1 (626) 395-3898 (office) Caltech, Broad Center, M/C 114-96
> +1 (626) 395-2000 (fax) 1200 E. California Blvd
> [log in to unmask] Pasadena, CA 91125
>
>
hank
|