There should be documentation discussing naming conventions and the package structure itself, some common features are listed right below:

• rename files to remove whitespaces, see #4
• rename files to make names short and concise
• move tests to a separate folder for clear distinction
• drop Stata files: see 4ed759b
• expand Readme
• check the license
• build a package, add fixed versions of dependencies

Make a new variable with binary married/cp/ch vs not outcome.

https://github.com/andrewbaxter439/UC-DiD-analysis/blob/38c86c2b490e6b83ced5e9335056e8f78588cd59/R/prediction_data_import.R#L116-L120

Use of QUAL_[1-35] excludes non-UK qualifications. Change to use HIQUL11D/HIQUL15D instead

https://github.com/andrewbaxter439/UC-DiD-analysis/blob/38c86c2b490e6b83ced5e9335056e8f78588cd59/R/prediction_data_import.R#L219-L226

I tried to run the project on the workstation using the data you sent me. It fails with this message:

Error in `group_by()`:
! Must group by variables found in `.data`.
✖ Column `benefit` is not found.
Backtrace:
    ▆
 1. ├─ggplot2::ggplot(...)
 2. ├─dplyr::arrange(...)
 3. ├─dplyr::filter(...)
 4. ├─dplyr::mutate(...)
 5. ├─dplyr::summarise(...)
 6. ├─dplyr::group_by(...)
 7. └─dplyr:::group_by.data.frame(...)
 8.   └─dplyr::group_by_prepare(.data, ..., .add = .add, caller_env = caller_env())
 9.     └─rlang::abort(bullets, call = error_call)
Execution halted

What do you think went wrong?

Since we run everything in parallel im might be a good idea to employ installed GPUs.
R provides some opportunities for that, however, I'm not sure your current code is 100% compatible with that.
That should theoretically make your calculations blazingly fast. Not sure about the accuracy though, single precision floating-point numbers can be pretty nasty.

Hi @vkhodygo - have updated the code in the optimise_predictions.R file to run a full set of optimising predictions. Will be a little longer than trial versions. Could you run then send the .rds files from the output folder (via email or onedrive)?

Most things fixed and should run smoothly, but let me know if not.