Alison Appling

Is it possible your compilation tools aren't up to snuff? Looks like it can't find C:/rtools42/mingw64/bin/g++ (a compiler). Start at the [rstan wiki page](https://github.com/stan-dev/rstan/wiki) for the most up-to-date rstan installation...

Haven't seen the issue in a while. Might not even need the requireNamespace call now that deSolve is an official Depends package, but leaving as-is because it isn't currently broken

Interesting. Does it help to call `library(desolve); library(streamMetabolizer)` as your very first calls in a fresh R session (with no automatically loaded data)?

Hi Moritz, > how much does the time window size (=number of consecutive days) affect the estimates? _consecutive_ days don't matter, but when you model a greater total number of...

> I actually would not expect K600 to vary all that much. it has been shown that the ponds are relatively stratified, with only very little mixing from wind etc.....

Hi Tzu-Yao, take out the `lapply(names)` from your `get_fit()` command.

I think `DO_R2` may indeed be what you're looking for - check out the formula at https://github.com/USGS-R/streamMetabolizer/blob/master/inst/models/b_np_oipi_tr_psrckm.stan#L142 and note that you probably need to add it to your `params_out` specification...

`err_proc_iid_sigma_scale` is the only/primary lever for affecting the magnitude of the fitted process errors. I'd like @robohall 's thoughts on this too, but I think we see the models appear...

Try updating the package with ```r remotes::install_github('USGS-R/streamMetabolizer') ``` (because `DO_R2` is a relatively new parameter)

Ah, thanks for the reminder - amazing how much I can forget after writing a whole paper on something! The `predict_DO()` function should return a data.frame with a column called...