Multiple Regression and one-sample t-test will produce the same
results. SPM only uses the GLM. The difference between labels in SPM
is how the user enters information about the design. As long as the
columns are the same in both models and the variance structure is the
same, you will get the same result.
Best Regards, Donald McLaren
=================
D.G. McLaren, Ph.D.
Research Fellow, Department of Neurology, Massachusetts General Hospital and
Harvard Medical School
Postdoctoral Research Fellow, GRECC, Bedford VA
Website: http://www.martinos.org/~mclaren
Office: (773) 406-2464
=====================
This e-mail contains CONFIDENTIAL INFORMATION which may contain PROTECTED
HEALTHCARE INFORMATION and may also be LEGALLY PRIVILEGED and which is
intended only for the use of the individual or entity named above. If the
reader of the e-mail is not the intended recipient or the employee or agent
responsible for delivering it to the intended recipient, you are hereby
notified that you are in possession of confidential and privileged
information. Any unauthorized use, disclosure, copying or the taking of any
action in reliance on the contents of this information is strictly
prohibited and may be unlawful. If you have received this e-mail
unintentionally, please immediately notify the sender via telephone at (773)
406-2464 or email.
On Tue, Feb 26, 2013 at 11:25 AM, Adam Gorka <[log in to unmask]> wrote:
> Hey Marco,
>
> Generally the correlations you find within SPM should also be present when
> tested outside SPM. My guess is that what you were testing within SPM was
> not what you wanted to test. Generally, when people perform 1 sample T tests
> with a covariate, they seek to ensure that the main effect of the contrast
> image are significant after controlling for another factor. I am not sure
> what you would be modeling by testing for effects of a covariate in a one
> sample T test, as I have never heard of someone looking for correlations
> between BOLD changes and another variable that way. It may be modeling BOLD
> changes after parceling out variability associated with the main effects of
> task.
>
> Again, I am not sure what a 1 sample T test with a covariate would be
> modeling if you test for effects of the covariate. But if you model it as a
> multiple regression, testing for a regressor will result in clusters that if
> you extract the parameters those parameters will be correlated with your
> regressor.
>
> Adam Gorka
>
>
>
>
> On Tue, Feb 26, 2013 at 10:29 AM, Marco Sperduti <[log in to unmask]>
> wrote:
>>
>> Thank you Gorka,
>>
>> I will try that...but do you have any clues about why I found different
>> results between SPM correlations and correlation computed outside SPM?
>>
>> sincerely,
>>
>> Marco
>>
>>
>> ________________________________
>> Da: Adam Gorka <[log in to unmask]>
>> A: [log in to unmask]
>> Inviato: Marted́ 26 Febbraio 2013 15:57
>> Oggetto: Re: [SPM] correlation problem
>>
>> Hello Marco,
>>
>> Are you modeling a 1 sample T test with neuropsychological score X as a
>> covariate?
>>
>> If so, I would recommend modeling the data with the "Multiple Regression"
>> model instead. Just enter the contrast images, and include
>> neuropsychological score X as a covariate.
>>
>> Hope this helps,
>>
>> Adam Gorka
>>
>>
>>
>> On Tue, Feb 26, 2013 at 4:51 AM, Marco Sperduti <[log in to unmask]>
>> wrote:
>>
>> Dear SPMers,
>>
>> I run a second level t-test using first level contrasts between my
>> condition of interest A and a rest condition. I found, at group level,
>> activation in several regions that choerent with my experimental task.
>>
>> Now, I want to test correlation between activation in these regions and a
>> neuropsychological score X.
>>
>> I've made a second model entering the images of the first level contrast
>> for each subject and X as a covariate. So I have a model with two columns,
>> first column=contrast images; second columns=X. Then to test for correlation
>> I used the contrast vector 0 1, and indeed i found 4 cluster showing
>> correlation in this way.
>>
>> The problem arise when I tried to extract signal from the four clusters
>> and run the same correlation outside SPM. I used MARSBAR to exctract the
>> signal from the four clusters, then I simply entered the values of the
>> signal for each cluster and X in a statistic software and looked for
>> correlations between these variables. In this way I didn't find any
>> correlation (the correlation are really far to be significant in this way).
>>
>> How is it possible? I've done something wrong? Is there another way to
>> test for correlation?
>>
>> Any help would be really appreciated.
>>
>> Kindly,
>>
>> Marco
>>
>>
>>
>>
>
|