Hi Gang,
I am trying to use 1dSVAR to perform functional connectivity analysis between 4 ROIs with 6 conditions as covariates. Several problems have came out:
1. Highest order (or number of lags) for the VAR model (1,2,...)?
[1] "Suggested orders for VAR:"
AIC(n) HQ(n) SC(n) FPE(n)
7 5 4 7
[1] "AIC: Akaike Information Criterion;"
[1] "HQ: Hannan-Quinn criterion;"
[1] "SC: Schwartz Criterion;"
[1] "FPE: Final Prediction Error criterion."
what does "order" mean here? lags between ROIs? what is the optimal number for fMRI data?
2. Now validating the instantaneous effects network...
Provide a specification matrix for the instantaneous effects.
ROI names should be provided for both column and row.
The diagonals should be 0, and each of the off-diagonals
indicates a connection from row to column, specified with
either 0 (not connected), NA (to be estimated), or a known
value (prior knowledge of path strength).
No. 1 specification matrix for instantaneous effects file name: xxx.1D
In this matrix file, if I believe region A is connecting with region B (but I don't know the path strength), should I input value 1 or NA (to be estimated)?
3. For the next step, it failed no matter whether I chose 0 (iterative) or 1 (direct)
The following two options are for numerical algorithm. Choose first
'Iterative', switch to 'Direct' if the iterative approach fails.
Numerical scheme (0: Iterative; 1: Direct)? 0
Error in `[<-`(`*tmp*`, pos
, i, value = 1) : subscript out of bounds
Numerical scheme (0: Iterative; 1: Direct)? 1
Error in SVAR(x = fm, estmethod = "direct", Amat = instMat, Bmat = bb, :
Model is not identified,
choose different settings for 'Amat' and/or 'Bmat'.
4. this question might be stupid, but if the program failed somewhere, it wouldn't let me redo the last step and continue, it seems like I have to restart from the very beginning! Is there some way I can save the calculated results and resume the program from where I failed?
Thanks
Yi