Dear Matthew:
At 08:21 PM 9/21/2005, Maxwell Boakye wrote:
>Darren
>thanks, comments have been really helpful
>
>here is a sample of values from a marsbar analysis I did of a left hand
>movement task.
>
>P corrected p ROI contrast value t satistic
>.00000 .000000 precentral R. 0.5 7.05
>
>P corrected p ROI contrast value t satistic
>.000111 .0001105 postcentral R. 0.24 3.77
>
>can i interpret this to mean the there was more activation in postcentral
>than precentral and can i use the contrast values of .5 and 0.24 for
>further stats ie can i take an average of all prcentral contrast values
>from 8 subjects who did the task and comapre that to average of
>postcentral values? or can i do correlational stuies between the precentral
>contrast values and the postcentral contrast values.? i.e
>is the contrast value the same as the measure of effect size- i.e can I use
>that as measure of strength to compare tasks and in serial studies
The contrast values should be c' * beta where c is the contrast vector
(e.g., [1 0 0 -1 ...]) and beta are the parameter estimates across the
design matrix. So yes I believe you can take the contrast values, in this
case one from each subject for each region and examine this at the second
level. Remember, because the pre and post-central gyrus contrast values are
taken from the same subjects you should use a paired t-test or you can
calculate the difference within the subject (0.5 - 0.24) and forward this
value to the second level. I believe you could also do a correlation on
these values; as you describe it above it would be across subjects. This is
similar to what used to be done in PET studies.
Darren
>thanks
>
>
>Quoting Darren Gitelman <[log in to unmask]>:
>
> > Hi Maxwell: Sorry this message got buried.
> >
> >
> > At 09:18 AM 9/16/2005, you wrote:
> > >Hi Darren
> > >
> > >Can you help me with this:
> > >
> > >I get a list of p values, corrected p values and t values after ROI
> > >analysis(using Marsbar). Sometimes the p values and corrected p values
> > are
> > >greater than .05, sometimes as high as 1.0
> > >Can I reject from list activated ROI's with p values greater than .05 or
> > >corrected p values greater than .05.
> >
> > I'm not sure what you mean by reject from list? Whether or not you use an
> > ROI a p value > 0.05 is not "significant." Whether or not to report it
> > anyway depends on the context and what you are trying to say.
> >
> > >Also for statistical analysis can I
> > >use the t value as a measurement of strength of activation for
> > comparison
> > >in serial studies?
> >
> > No. t-values are measures of reliability not strength of activation.
> > remember t-value formula is the mean / standard error. The strength of
> > activation can either be calculated as a percent signal change (see
> > Matthew
> > Brett's notes on this on the marsbar page) or by using the parameter
> > estimates (measure of effect size).
> >
> >
> > > Does the same apply to the SPM analysis and the values
> > >generated by Volume and Cluster buttons?
> >
> > Yes.
> >
> > >Lastly it occasionally gives negative t values ( the p values for these
> > are
> > >usually high)-can I reject these?
> >
> > Negative t-values are perfectly reasonable but just says the effect is in
> > the opposite direction. Whether to reject these is based on what you are
> > examining.
> >
> > Regards,
> > Darren
> >
> > >thanks
> > >
> > >--
> > >Maxwell Boakye, M.D.
> > >Assistant Professor
> > >Department of Neurosurgery
> > >Stanford University Medical Center
> > >300 Pasteur Dr
> > >Stanford, CA, 94305-5327
> > >Phone: 650-493-5000,64313
> > >Fax: 650-849-0237
> >
> >
> > -------------------------------------------------------------------------
> > Darren R. Gitelman, M.D.
> > Cognitive Neurology and Alzheimer¹s Disease Center
> > Northwestern Univ., 320 E. Superior St., Searle 11-470, Chicago, IL 60611
> > Voice: (312) 908-9023 Fax: (312) 908-8789
> > -------------------------------------------------------------------------
> >
> >
> >
>
>
>--
>Maxwell Boakye, M.D.
>Assistant Professor
>Department of Neurosurgery
>Stanford University Medical Center
>300 Pasteur Dr
>Stanford, CA, 94305-5327
>Phone: 650-493-5000,64313
>Fax: 650-849-0237
-------------------------------------------------------------------------
Darren R. Gitelman, M.D.
Cognitive Neurology and Alzheimer¹s Disease Center
Northwestern Univ., 320 E. Superior St., Searle 11-470, Chicago, IL 60611
Voice: (312) 908-9023 Fax: (312) 908-8789
-------------------------------------------------------------------------
|