MaximilianPi
commented
5 days ago Hi @billurbektas,
Thanks for your interest in sjSDM.
Yes, there was no support for family = gaussian() until now. However, I just pushed an update to the development version (basically the github version). You will need to reinstall sjSDM by running
devtools::install_github("https://github.com/TheoreticalEcology/s-jSDM", subdir = "sjSDM", ref = "master")
It has not been thoroughly tested yet, but some initial checks indicate reliable results. Feel free to report here if you see anything suspicious.
PS: You do not have to re-install the python setup, only the R package must be updated to the github/development version
billurbektas
Hi, I was exploring the option to fit a gaussian distribution to my abundance data (324 sites x 600 species). The model runs, however later anova and Rsquared functions are giving an error. I was wondering if this an error from my part or it is just that the gaussian fit is not fully implemented yet as stated in the error. Thank you very much in advance for your support!
mod.cal = sjSDM(Y = as.matrix(veg.abund), env = linear(data = veg.env, formula = ~MAT + AP), spatial = linear(data = veg.env, formula = ~0+x*y), family = gaussian("identity"), se = TRUE, verbose = TRUE, iter = 20)
Rsquared(mod.cal, verbose = FALSE) #0.01 Error in py_call_impl(callable, call_args$unnamed, call_args$named) : TypeError: add() received an invalid combination of arguments - got unrecognized keyword arguments: dtype, device Run
reticulate::py_last_error()for details. In addition: Warning message: In get_null_ll(model, verbose = verbose) : family = gaussian() is not fully supported yet.