New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Issue 289: Getting help with stan vignette #405
Conversation
Merge branch 'main' of https://github.com/epinowcast/epinowcast into stan-help
…stan` syntax. (#386) * first draft action that installs cmdstan * use the correct approach to checking for stan changes * add check for cmdstanr compilation and syntax correctness * add skeleton news item * update PR number * use pull request not PR * Automatic readme update [ci skip] * make sure to install epinowcast * missing testthat call * Automatic readme update [ci skip] * remove spurious readme figures * add comments for cmdstan workflow * make the refex match exactly only * try a simpler approach to error catching * fail around more --------- Co-authored-by: GitHub Action <action@github.com>
Tagging @zsusswein, too. I wanted to get this in front of everyone. Happy for feedback (i.e., take a machete to it). I'm happy to edit/expand/etc after feedback. I feel like it needs some examples using synthetic data which I can make (e.g., put in too little data and let the model fail, put in multi-modal data and let the posteriors be wide and ESS low). |
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## main #405 +/- ##
=======================================
Coverage 96.84% 96.84%
=======================================
Files 15 15
Lines 1873 1873
=======================================
Hits 1814 1814
Misses 59 59 ☔ View full report in Codecov by Sentry. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is so good!
A few comments mostly discussing my personal quest for validation via control of spacing. I think we could link out to more detail on some of these topics (like adapt delta settings etc) a bit more liberally and we might also want to give some tips for working with cmdstan objects (like tidybayes
, posterior
etc.).
I think we should aim to not make this PR review into what we would ultimately like in this vignette but instead aim for a good skeleton we can add to over time.
I feel like it needs some examples using synthetic data which I can make (e.g., put in too little data and let the model fail, put in multi-modal data and let the posteriors be wide and ESS low).
I think this would be lovely but would also be very happy with a first version without this and save adding for a later PR?
Co-authored-by: Sam Abbott <s.e.abbott12@gmail.com>
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is really nice! I'm new to {epinowcast}
myself and found this a helpful guide. I left a bunch of comments inline -- they're split between minor typos and some suggestions around content to expand on or re-organize. I would consider them all optional and nitpicky.
One other thought: I would personally find it useful if this document mentioned some particular parameters to investigate in problematic fits. I don't know what the hierarchical variance parameters are called (or other frequent trouble spots), but it would speed up my model debugging to know where to look first. That may be out of scope, but thought it was worth mentioning.
Nice comments @zsusswein.
I think we should make this out of scope for this PR but open an issue to add to this vignette (or possibly to expand into more than one if getting long) as I agree it would be very useful. Part of the reason I think its out of scope is I think this might need #349 to be completed first otherwise its quite reliant on folk wisdom. |
@medewitt would it be helpful to add some suggestions for any of the comments? I think all/most of the reviewers would be happy to help so just give them/me a ping if useful! |
Sure! Sorry, I got a bit busier this weekend than planned. I should be able to pick this up Wednesday/Thursday this week.... |
Co-authored-by: Carl A. B. Pearson <pearsonca@users.noreply.github.com>
Co-authored-by: Zachary Susswein <46581799+zsusswein@users.noreply.github.com>
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This looking even better now. A few more comments.
@seabbs almost there. I'll fix the linting error ( If you feel good about it, I'll add a news item and move it out of draft. |
Awesome - I do indeed feel good about it! |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is really nice. Thanks for putting so much effort into it @medewitt! I think you've addressed all the reviewer points (thanks @zsusswein and @pearsonca!) so this is good to go from my perspective.
Note I adjusted the pkgdown articles section to split out help with epinowcast from case studies as I think this a little bit clearer.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is really nice. Thanks for putting so much effort into it @medewitt! I think you've addressed all the reviewer points (thanks @zsusswein and @pearsonca!) so this is good to go from my perspective.
Note I adjusted the pkgdown articles section to split out help with epinowcast from case studies as I think this a little bit clearer.
Description
Getting the discussion on the PR to draft mode for future discussion. This PR closes #289.
This PR includes the initial draft for users getting help with Stan. It provides some extended information on the following:
epinowcast
functionsChecklist