sdmTMB
sdmTMB copied to clipboard
:earth_americas: An R package for spatial and spatiotemporal GLMMs with TMB
I followed [these](https://stackoverflow.com/a/70664229) instructions (M1 Mac specific, I think). Then worked on [this](https://github.com/pbs-assess/sdmTMB/tree/parallel) branch. Tested in [this](https://github.com/pbs-assess/sdmTMB/blob/parallel/inst/test-open-mp.R) script. Depends strongly on the model, but I'm getting ~50% speed ups up...
Initial development in CAR branch: To do list: - spatial and spatiotemporal components currently included, with the latter as IID. Does it make sense to also include AR(1)? - matrix.inverse()...
Just ran across the sigma() function in glmmTMB. Seems useful for nbinom2, etc. Also in lme4 https://github.com/glmmTMB/glmmTMB/blob/41a8abcd99692d6fe9e7d82f454e4ec57ab3f162/glmmTMB/R/VarCorr.R#L87
The majority of this is now done in the [dispformula](https://github.com/pbs-assess/sdmTMB/tree/dispformula) branch. See below. Still todo: - [ ] add to `tidy.sdmTMB()` - [ ] make sure simulation interfaces work -...
The model likely needs to be tweaked then to return all random effects for normalize = true.
See mgcv for inspiration. Lets you omit some predictors from the newdata. Also, in mgcv, they have an additional argument in predict() 'newdata.guaranteed' which if TRUE results in the function...
@Kotkot made the suggestion and provided some example code. > Imagine we have rand1, rand2, and rand3 random effects acting on the intercept in the model. Each have 10, 20,...
Hi, I'm trying to make a prediction for validation ( using 20% of the original dataset) from a fitted ( using 80% of the original dataset) GLM model with spatial-temporal...
Currently silently doing nothing. Offset is meant to be via `offset` argument.