BrainVoyager Discussion Forum
  Statistics
  Supra-additive effect

Post New Topic  Post A Reply
profile | register | preferences | faq | search

UBBFriend: Email This Page to Someone! next newest topic | next oldest topic
Author Topic:   Supra-additive effect
DianaR
Junior Member
posted 15 May 2006 15:44     Click Here to See the Profile for DianaR   Click Here to Email DianaR     Edit/Delete Message   Reply w/Quote
hi, I'm having trouble figuring out how to calculate a supra-additive effect, meaning that activation to condition A is greater in a specific region than the sum of conditions B and C.

For example, in the BV sample with LVF, RVF, and BVF (bilateral VF), imagine that blocks of LVF, RVF, and BVF were collected with rest in between blocks. My question is whether there are any regions that show more activity to BVF than to the sum of LVF and RVF -- indicating that BVF is supra-additive, more than the sum of the two parts.

How do I do this analysis (I have 10 subjects, 3 runs per subject)?

Fabri
Moderator
posted 16 May 2006 14:40     Click Here to See the Profile for Fabri   Click Here to Email Fabri     Edit/Delete Message   Reply w/Quote
I see no special problem in running the desired GLM contrast.
You just have to specify in the Overlay GLM dialog, by checking condition A with + and conditions B and C with -. This should be exactly what you are looking for, namely comparing the condition A to the sum of condition B and C effects.

Many regards,
Fabrizio

quote:
Originally posted by DianaR:
hi, I'm having trouble figuring out how to calculate a supra-additive effect, meaning that activation to condition A is greater in a specific region than the sum of conditions B and C.

For example, in the BV sample with LVF, RVF, and BVF (bilateral VF), imagine that blocks of LVF, RVF, and BVF were collected with rest in between blocks. My question is whether there are any regions that show more activity to BVF than to the sum of LVF and RVF -- indicating that BVF is supra-additive, more than the sum of the two parts.

How do I do this analysis (I have 10 subjects, 3 runs per subject)?


DianaR
Junior Member
posted 16 May 2006 15:23     Click Here to See the Profile for DianaR   Click Here to Email DianaR     Edit/Delete Message   Reply w/Quote
Hi Fabrizio,

I've been thinking about this more, and I can explain better with a specific example:

I know that if I do the following in overlay GLM:
+ AV
- A
- V
and balance, the weights are 2, -1, -1
If I do a random effects analysis (10 ss, 3 runs each), at t=4.8, p<.001
I find a cluster of activation in the right STS of 8791 voxels to the AV stimulus. Other than bilateral STS, there are no other large activations showing AV > average (A, V)

If I do not balance the equation, the weights are 1, -1, -1
I thought this might be the additive approach I am looking for, since it doesn't balance the AV against the 2 unimodal conditions. The right STS activation is smaller at the same t value (2859 voxels), which initially made me think I was correct in interpreting this as a supra-additive effect (the AV effect is greater than the sum of the 2 unimodal effects, suggesting integration of the 2 modalities), but in this analysis there are a number of other activations (thalamus, retrosplenial cortext, visual cortex) and that doesn't make sense to me. If this is truly additive, there should be fewer active regions compared to the previous analysis, not more. AV against the sum of A and V should produce less activation than AV against .5A and .5V.

Am I interpreting this correctly? If not, can you please direct me in conducting this analysis?

Rainer Goebel
Administrator
posted 17 May 2006 10:59     Click Here to See the Profile for Rainer Goebel   Click Here to Email Rainer Goebel     Edit/Delete Message   Reply w/Quote

We tried such an analysis using conjunction of contrasts as described in the paper "Van Atteveldt, N., Formisano, E., Goebel, R., Blomert, L. (2004). Integration of letters and speech sounds in the human brain. Neuron, 43, 271-282. Maybe you want to look at that paper for details. The approach suggested by Fabri ("+AV" with "-A -V") is fine if used unbalanced as you correctly note. When balancing such a contrast, you get a comparison of AV with the mean of A and V. A conjunction of contrasts might be better to sort out irrelevant areas, i.e. "(AV > A) AND (AV > V) AND (A > 0) AND (V > 0)". While this does not test for superadditivity, it masks out the relevant voxels, which can then be tested for nonlinearity (sub/superadditivity).

If you use the single, unbalanced contrast approach, you get everything active with regard to a difference without ensuring that your "right" side ("A", "V") is also activated - it could well be that the right side effects ("A" and/or "B") are zero or negative.

Note also that true superadditivity is difficult to find in fMRI because of saturation effects, both at the neuronal and the hemodynamic level.

All times are ET (US)

next newest topic | next oldest topic

Administrative Options: Close Topic | Archive/Move | Delete Topic
Post New Topic  Post A Reply
Hop to:

Contact Us | BrainVoyager.com

Copyright Rainer Goebel 2001 - 2006

Powered by Infopop www.infopop.com © 2000
Ultimate Bulletin Board 5.47d