Closed ejardim closed 8 years ago
It seems very difficult because it must be set on the coefficients of the model, which do not have a simple domain. It could be possible by using the design matrix to compute the boundaries in the coefficients space ?!
Tried a couple of options, it doesn't seem to have a simple solution. We'd need a generic solution for the bounds and initial values, which is not feasible, it depends on the submodels used. A ~year requires completely different coefficients than ~factor(year), for example.
Boudaries in Q log(]0,1]) and F log(]0,2]) shouldn't be too difficult to include and I guess they will help a lot the minimization.