3dNLfim help


I am trying to use 3dNLfim to perform nonlinear least square fitting on the following model:

Y[t] = AX[t] + BZ[t].

I know Y[t], X[t] and Z[t], and I want to estimate A and B.
I was thinking to use 3dNLfim Expr2 model, but I am not sure how to set up the command. Any suggestion is very much appreciated.

Thank you,

Hi Giuseppe,

That looks like a linear model that could be solved with just 3dDeconvolve. Are either X or Z voxelwise? Presumably Y is, is that right?

  • rick

Thank you very much for your help Rick.

Y is voxel-wise, but X and Z have single values for each time point.


Given that, you should be able to run this using 3dDeconvolve, passing X and Z via -stim_file, and you can ponder whether -polort is appropriate and such.

Does that seem reasonable? I would be happy to ponder further.

  • rick

Thank you Rick.
3dDeconvolve worked, but the results for A and B were off compared to what I was expecting.
I noted that I could further simplify my equation by dividing both Y and X by Z, so that I ended up with only 2 variables (Y/Z and X/Z) and I could then run 3dNLfim. This led to results more reasonable and similar to what I was expecting.

Thank you again!