Regression - Generalized Additive Model?

Share icon

Wed, 08/30/2017 - 02:42 am

Dear all,

I am not quite confident with all the mathematics beneath regression, but is the use Generalized Additive Model fitting is available in Igor?
For instance, I'd like to fit C (which a 2Dwave dependent of u and v), so that sqrt(C) = b + s(u,v) + e
where:
- b is the overall mean of the response
- s(u,v) the isotropic smooth function of covariate u & v
- e is the residual

Thanks for your help !
Best
J-E
I really think that you should re-title your post; you are posing a fitting problem that does not immediately lend itself to any form of interpolation as implemented in ImageInterpolate unless you can find your desired expression to match the kriging variogram model choices under /K flag.

August 30, 2017 at 05:55 am - Permalink

Right. Told you I wasn't quite confident with all the interpolation/regression thing :D
Sorry for the confusion.
Pretty sure now that what I am trying to do is a regression. Does it get any better?

I first tried to play with Kriging and the parameters of /K but I am not understanding everything...

August 30, 2017 at 06:42 am - Permalink

Can you explain in more detail what your equation means. If e is the residual then you would normally not include it. In a regular fit the goal is usually to determine b and s(u,v). In kriging you do not get that information; you only get a smooth surface. It would be helpful if you could explain exactly what data you have and what you need to compute.

AG

August 30, 2017 at 07:56 am - Permalink