I had problems installing rstan (2.18.2) under R 3.6.0 on a Windows 10 machine by following the default instructions.
The problem was with Rtools3.5. Although the line:
pkgbuild::has_build_tools(debug = TRUE)
returned TRUE as it should, the tool chain was not working.
After some foul language, some googling and a few hours lost I found and fixed the problem. In order to save other people from losing time I thought I just post my solution in one place.
Unlike indicated in the installing instructions, Rtools needs to be included in the system path like so:
Sorry for the headache and thanks for sharing the fix that worked for you.
I don’t have Windows so can’t check for myself, maybe @bgoodri or @paul.buerkner can confirm? We definitely need to change the instructions if this is not some weird special case.
Had the exact same problem - solved it for me as well. Many thanks, Seb.
It took me HOURS to get to a solution - what is the problem with R 3.6, Rtools, and Stan not playing well with each other? I’d like to try and avoid these issues at work in the future, as it caused considerable downtime for my team.
I’ve been fighting with conflicts resulting from updating to R 3.6 for days now. Specifically, when fitting models using brms that include group variables, my R just dies without a message. rstanarm functions seem to work fine, though.
Just tried this, and it still isn’t working properly for me, unfortunately. It seems a lot of people are having this 3.6/rstan/Windows 10 conflict.