[Freesurfer] question about weighted regression analysis

[Katie Bettencourt]

So I created a weighted regression analysis to look at the effect of memory
load in a particular brain region. Basically, I weighted the paradigms by a
behavioral measure that reflected the number of items actually remembered
(as set size was increased).  As far as Doug told me there are basically 2
ways to weight your paradigm files.

Version 1:
Have 2 conditions, baseline (condition 0) and all the set sizes (condition
1).  Condition 1 would then be weighted by the behavioral measure.

Version 2:
Have 3 conditions, baseline (condition 0), and then I represented each
presentation as two different conditions, one with a weight that is always 1
(condition 1), the other weighted according to the behavioral measure
(condition 2).


The difference, as far as I understand it, in version 1, it is assumed that
the response amplitude is ) when the weight is 0 (ie. that when you are
attending to 0 items, brain activity = 0).  Whereas, version 2, tests the
slope of the HRF amplitude vs weight without the assumption above.

However, I'm a bit confused as to the results I got.  When I looked at the
data from both versions, version 1 provided a much higher amount of
activation and more areas activated than version 2.  However, I believe
version 2 better fits with the multiple regression analysis that is done in
Brain Voyager.

Can anyone give me a better explanation of what the difference between these
analysis models is?

Katie
___
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer


The information in this e-mail is intended only for the person to whom it is
addressed. If you believe this e-mail was sent to you in error and the e-mail
contains patient information, please contact the Partners Compliance HelpLine at
http://www.partners.org/complianceline . If the e-mail was sent to you in error
but does not contain patient information, please contact the sender and properly
dispose of the e-mail.

Re: [Freesurfer] question about weighted regression analysis

[Douglas N Greve]

I assume that you are comparing maps of Version1:1v0 and Version2:2v0 ? 
I could imagine it going either way. If the true slope is 0 but the 
offset is non-0, then Version1 will give you an artificially high slope 
(and Verion2 will give you the correct slope at 0, and so no 
activation). Are you comparing this to a BV analysis?

doug

Katie Bettencourt wrote:
> So I created a weighted regression analysis to look at the effect of 
> memory load in a particular brain region. Basically, I weighted the 
> paradigms by a behavioral measure that reflected the number of items 
> actually remembered (as set size was increased).  As far as Doug told 
> me there are basically 2 ways to weight your paradigm files.
>
> Version 1:
> Have 2 conditions, baseline (condition 0) and all the set sizes 
> (condition 1).  Condition 1 would then be weighted by the behavioral 
> measure.
>
> Version 2:
> Have 3 conditions, baseline (condition 0), and then I represented each 
> presentation as two different conditions, one with a weight that is 
> always 1 (condition 1), the other weighted according to the behavioral 
> measure (condition 2).
>
>
> The difference, as far as I understand it, in version 1, it is assumed 
> that the response amplitude is ) when the weight is 0 (ie. that when 
> you are attending to 0 items, brain activity = 0).  Whereas, version 
> 2, tests the slope of the HRF amplitude vs weight without the 
> assumption above.
>
> However, I'm a bit confused as to the results I got.  When I looked at 
> the data from both versions, version 1 provided a much higher amount 
> of activation and more areas activated than version 2.  However, I 
> believe version 2 better fits with the multiple regression analysis 
> that is done in Brain Voyager. 
>
> Can anyone give me a better explanation of what the difference between 
> these analysis models is?
>
> Katie
>
>
>

-- 
Douglas N. Greve, Ph.D.
MGH-NMR Center
gr...@nmr.mgh.harvard.edu
Phone Number: 617-724-2358 
Fax: 617-726-7422

Bugs: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting
FileDrop: www.nmr.mgh.harvard.edu/facility/filedrop/index.html

___
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer


The information in this e-mail is intended only for the person to whom it is
addressed. If you believe this e-mail was sent to you in error and the e-mail
contains patient information, please contact the Partners Compliance HelpLine at
http://www.partners.org/complianceline . If the e-mail was sent to you in error
but does not contain patient information, please contact the sender and properly
dispose of the e-mail.

Re: [Freesurfer] question about weighted regression analysis

[Douglas N Greve]

You should also look at Version2:1v0. I bet a lot of the areas from 
Version1:1v0 will also show up. You can also create a Version3 in which 
you divide your presentations into a low-weight and a high-weight (but 
set the weight=1). Then create contrasts of low+high and high-low. The 
low+high should look like Version2:1v0 and the high-low should look like 
Version2:2v0.

doug




Katie Bettencourt wrote:
> Yes, those are the maps I"ve been comparing.  I've been comparing it 
> to BV sort of, but that analysis is not surface based and I"m not used 
> to it, so I can't quite tell which is more accurate, though Version 2 
> gives a much smaller area of activity, which fits with the description 
> of what I've been given about what to expect in BV.  Attached is two 
> pictures of the difference I get for Version 1:1v0 (labeled with 
> "single" in the image name) and Version 2:2v0 (labeled with "double" 
> in the name).  As you can see, Version1 activates a much larger area 
> than Version 2.
>
> I guess part of my problem is that I'm having trouble understanding 
> exactly what these two versions are telling me about the data and what 
> the differences is.  Can you try to give me a sort of layman's 
> explanation?
>
> Katie
>
>
> On Tue, Feb 22, 2011 at 11:45 AM, Douglas N Greve 
> mailto:gr...@nmr.mgh.harvard.edu>> wrote:
>
> I assume that you are comparing maps of Version1:1v0 and
> Version2:2v0 ? I could imagine it going either way. If the true
> slope is 0 but the offset is non-0, then Version1 will give you an
> artificially high slope (and Verion2 will give you the correct
> slope at 0, and so no activation). Are you comparing this to a BV
> analysis?
>
> doug
>
>
> Katie Bettencourt wrote:
>
> So I created a weighted regression analysis to look at the
> effect of memory load in a particular brain region. Basically,
> I weighted the paradigms by a behavioral measure that
> reflected the number of items actually remembered (as set size
> was increased).  As far as Doug told me there are basically 2
> ways to weight your paradigm files.
>
> Version 1:
> Have 2 conditions, baseline (condition 0) and all the set
> sizes (condition 1).  Condition 1 would then be weighted by
> the behavioral measure.
>
> Version 2:
> Have 3 conditions, baseline (condition 0), and then I
> represented each presentation as two different conditions, one
> with a weight that is always 1 (condition 1), the other
> weighted according to the behavioral measure (condition 2).
>
>
> The difference, as far as I understand it, in version 1, it is
> assumed that the response amplitude is ) when the weight is 0
> (ie. that when you are attending to 0 items, brain activity =
> 0).  Whereas, version 2, tests the slope of the HRF amplitude
> vs weight without the assumption above.
>
> However, I'm a bit confused as to the results I got.  When I
> looked at the data from both versions, version 1 provided a
> much higher amount of activation and more areas activated than
> version 2.  However, I believe version 2 better fits with the
> multiple regression analysis that is done in Brain Voyager.
> Can anyone give me a better explanation of what the difference
> between these analysis models is?
>
> Katie
>
>
>
>
> -- 
> Douglas N. Greve, Ph.D.
> MGH-NMR Center
> gr...@nmr.mgh.harvard.edu 
> Phone Number: 617-724-2358 Fax: 617-726-7422
>
> Bugs: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting
> 
> FileDrop: www.nmr.mgh.harvard.edu/facility/filedrop/index.html
> 
>
>
>
> The information in this e-mail is intended only for the person to
> whom it is
> addressed. If you believe this e-mail was sent to you in error and
> the e-mail
> contains patient information, please contact the Partners
> Compliance HelpLine at
> http://www.partners.org/complianceline . If the e-mail was sent to
> you in error
> but does not contain patient information, please contact the
> sender and properly
> dispose of the e-mail.
>
>
>
> 
>
>
> 
>

-- 
Douglas N. Greve, Ph.D.
MGH-NMR Center
gr...@nmr.mgh.harvard.edu
Phone Number: 617-724-2358 
Fax: 617-726-7422

Bugs: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting
FileDrop: www.nmr.mgh.harvard.edu/facility/filedrop/index.html

___
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer