I am trying to run 3dLSS on a 1D timeseries data. The timeseries is of size 5MB, and the design matrix is of size 179.4 MB. When I try to run 3dLSS I get the following error:
0m malloc failure for image space: 2904614976 bytes
I also tried running it by requesting more memory, but it still kept giving me the same error. Is there a solution to this?
Chirag, why not directly using -stim_times_IM in 3dDeconvolve unless you have multicollinearity problem?
Yes, there is multicollinearity issue.
How severe is the multicollinearity problem with stim_times_IM? What basis function are you using?
Collinearity is very high. Several regressors have VIFs greater than 10. Whereas the model created by 3dLSS at each iteration has VIF less than 2.
Anyhow, 3dLSS did work after I downsampled the timeseries data from 250Hz to 10Hz.