hi, Stephen,
Thanks for your comments.
If my experiment was a simple block design, only included one session that was consisted of one baseline condition and one task condition, and my contrast was defined as the task-baseline. Thus the grand mean scaling should equal the local mean value, right? Then can i say that beta value represent the percent signal change associated with the task in my paper?
hengyi
----------------------------
>On Sat, 26 Mar 2005 11:40:53 -0500, [log in to unmask]
><[log in to unmask]> wrote:
>
>>hi, Darren,
>>
>>Thanks for your quickly reply. I know that Beta values do not directly
>represent the signal change. But many paper reported the beta (task) as
>the percent signal change associated with that task, as mentioned in one
>of Will's mail http://www.jiscmail.ac.uk/cgi-bin/webadmin?
>A2=ind04&L=spm&P=R392137&I=-1.
>
>What Will said in that post is that the betas are in *units* of percent
>signal change. That doesn't mean the betas truly represent *local*
>percent signal change. That's because the factor used in grand mean
>scaling is "'100/g' where g is the *average* value over all time points
>and scans in that session." [from Will's post; emphasis added]
>
>>Thus, i am wondering what reference i can cite if i also use the beta in
>my paper.
>
>If you used the SPM default of grand mean scaling, I think it's best not
>to call the beta "percent signal change," because people will read that as
>*local* signal change, not merely "units of".
>
>If you really want percent signal change, you should normalize each voxel
>by its own mean (over each run/session). Will goes over this in his post
>("If you wish to report in units of percent of 'local' signal, then
>because the average local signal is simply beta(constant term)...") (And
>if you're doing a random effects analysis, this should be done before
>passing to the second (group) level.)
>
>Best,
>
>S
>
>>
>>thanks and best,
>>
>>
>>hengyi
>>----------------------------
>>
>>
>>>Dear Hengyi:
>>>
>>>Beta values do not represent signal change, at least not in the direct
>>>sense. Please look at Matthew Brett's marsbar faq for a clear explanation
>>>of the relationship between beta's and percent signal change.
>>>
>>>http://marsbar.sourceforge.net/faq.html
>>>
>>>Darren
>>>
>>>At 03:19 PM 3/25/2005, [log in to unmask] wrote:
>>>>Dear SPM members,
>>>>
>>>>If i report beta value to represent the percent signal change associated
>>>>with the task in my paper, what references shoulf i cited?
>>>>
>>>>thanks and best
>>>>
>>>>----------------------------
>>>>Hengyi Rao, Ph.D.
>>>>[log in to unmask]
>>>>Department of Neurology
>>>>University of Pennsylvania
>>>>3 W Gates
>>>>3400 Spruce Street
>>>>Philadelphia, PA, 19104
>>>>----------------------------
>>>
>>>
>>>-------------------------------------------------------------------------
>>>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
>>>-------------------------------------------------------------------------
>>
>>= = = = = = = = = = = = = = = = = = = =
= = = = = = = = = = = = = = = = = = = =
|