help
help > RE: Weighted GLM correlation or regression
Jan 24, 2022 11:01 PM | Alfonso Nieto-Castanon - Boston University
RE: Weighted GLM correlation or regression
Dear Alex,
It is perfectly fine either way (using regression- or correlation- based measures in the context of gPPI analyses), typically gPPI analyses have used regression measures while non-gPPI analyses (e.g. weighted GLM) have used correlation measures, but that is mostly out of convention/tradition, and both approaches are perfectly fine/justified. Choosing regression or correlation measures does not change the manner in which the analysis is run, it only affects the "units" of the output/measured effects, where regression coefficients measure connectivity as "change in BOLD activation (in PSC units) in target area over change in BOLD activation (in PSC units) in source area" while correlation coefficients measure connectivity as "change in BOLD activation (in standard deviation units) in target area over change in BOLD activation (in standard deviation units) in source area" units
Hope this helps
Alfonso
Originally posted by alexandre :
It is perfectly fine either way (using regression- or correlation- based measures in the context of gPPI analyses), typically gPPI analyses have used regression measures while non-gPPI analyses (e.g. weighted GLM) have used correlation measures, but that is mostly out of convention/tradition, and both approaches are perfectly fine/justified. Choosing regression or correlation measures does not change the manner in which the analysis is run, it only affects the "units" of the output/measured effects, where regression coefficients measure connectivity as "change in BOLD activation (in PSC units) in target area over change in BOLD activation (in PSC units) in source area" while correlation coefficients measure connectivity as "change in BOLD activation (in standard deviation units) in target area over change in BOLD activation (in standard deviation units) in source area" units
Hope this helps
Alfonso
Originally posted by alexandre :
Dear all,
This is a subject that has already been discussed but the answer still escapes me...
While reading the handbook (https://www.researchgate.net/publication...) for SBC analyses, I saw that for gPPI, it was necessary to define a regression. However, I don't see the info for a weighted-GLM (analysis presented just before gPPI).
It seems to me that you have to choose 'correlation' for weighted-GLM, but I confess I don't really understand why...
Could someone explain it to me?
Thanks
Alex
This is a subject that has already been discussed but the answer still escapes me...
While reading the handbook (https://www.researchgate.net/publication...) for SBC analyses, I saw that for gPPI, it was necessary to define a regression. However, I don't see the info for a weighted-GLM (analysis presented just before gPPI).
It seems to me that you have to choose 'correlation' for weighted-GLM, but I confess I don't really understand why...
Could someone explain it to me?
Thanks
Alex
Threaded View
Title | Author | Date |
---|---|---|
alexandre | Jan 24, 2022 | |
Alfonso Nieto-Castanon | Jan 24, 2022 | |
alexandre | Jan 25, 2022 | |