search

mfasiolo / qgam

Additive quantile regression R package
http://mfasiolo.github.io/qgam/
30 stars 7 forks source link

Avoid gam.setup() in Extended Gam calibration #6

Closed mfasiolo closed 8 years ago

mfasiolo commented 8 years ago

Most of the time is currently spent by gam.setup() which creates the model.

mfasiolo commented 8 years ago

I was profiling the calibration code. In the extended gam version at least, 90% of the time is being spent by gam.setup(), which gets called for each value of sigma.

I would like to create a single gam object (G), and then change its value of sigma and its smoothing parameters (which are not estimated).

For instance:

G <- gam(y~s(x), family = logF(qu = qu, lam = lam, theta = 0), fit = FALSE, sp = fit1$sp)

The problem is that sigma is easy to reset:

G$family$putTheta( 3 )

But I don't know what to do to change sp, because I guess it's incorporated in the penalty matrices. At least sp is not here

G$sp named numeric(0)

If I try to put something in G$sp and then estimate it does not work!