Hi Jungwoo,
remember the SPM matrix also have the filter so pinv(SPM.Xx.X) won't
give the exact same results -- you could however specify a matrix with
basic motion parameters (expended or not) and filter and then work on
residuals for estimation (this assumes can stuill a bit different
beceause now you explain away motion 1st and then whatever regressors ie
the subspaces spanned are orthogonal)-- if inference adjust df/dfe
cyril
> Hello, SPM users!
>
> If I can specify every columns of design matrix, can I just use pinv function to get beta weights or residuals?
> In the case where I have my own elements in design matrix, it is burdensome to convert them into a form of matlabbatch.
>
> Actually, I checked the result myself, and found out that correlation of betas from each results were more than .95 but show some difference in their magnitude.
>
> So I thought it may due to some modification of the brain data itself, but I found no clue.
> Your answers would be of great help.
>
> Thank you!
>
> Best,
> Jungwoo Kim.
>
>
>
>
--
Dr Cyril Pernet,
Neurobiology Research Unit
Copenhagen University Hospital, Rigshospitalet
Building 8057, Blegdamsvej 9
DK-2100 Copenhagen, Denmark
[log in to unmask]
https://cpernet.github.io/
https://orcid.org/0000-0003-4010-4632
|