paper-comparison vignette

[ha0ye]

Summary of changes:

• no longer rely on files in the vignettes folder (i.e. the ~150MB christensenetal-comparison-files folder)
  • instead, download files from Erica's repo, the PortalData repo, and a separate replication repo
• use a single control parameter to dictate whether to run the models or not
  • this is a rmarkdown parameter set in the yaml header
  • by default, it is set to false, and so does not run the LDA or TS models, instead downloading them
  • when set to true, it foregoes downloading the model objects, and runs all the LDA and TS models
• consolidate duplicated code
  • for handling the various outputs from running the LDA model, running the TS model, selecting the best LDA and TS model, and getting the dates
• make the changepoint results consistent objects
  • save the whole posterior distributions for the changepoints (which simplifies follow-up analyses)
• some general cleanup of code chunks, section headers, etc.

closes #120

[juniperlsimonis]

are the build breaks here the random number thingy?

[ha0ye]

For the last commit, yes. (I haven’t looked at the logs for all prior commits).

[codecov-io]

Codecov Report

Merging #124 into master will not change coverage. The diff coverage is n/a.

@@           Coverage Diff           @@
##           master     #124   +/-   ##
=======================================
  Coverage   97.56%   97.56%           
=======================================
  Files          11       11           
  Lines        1191     1191           
=======================================
  Hits         1162     1162           
  Misses         29       29

---

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update bf918db...fca81b5. Read the comment docs.

[juniperlsimonis]

finally got back to this (sorry!) and with rebasing this branch with the updates you had in the other pr, @ha0ye it looks like that takes care of the test issue (woo hoo!), which was my concern here. @diazrenata do you have any editing feedback or can this be merged in?

[juniperlsimonis]

do note that we'll need to update some of the code and model objects with the jump to v0.2.0 (ref: #131 ), so i'm not sure which we want to do first?

[ha0ye]

do note that we'll need to update some of the code and model objects with the jump to v0.2.0 (ref: #131 ), so i'm not sure which we want to do first?

I think this PR is independent enough (it only changes 2 files) that it can be rebased on top of 0.2.0; and the vignette should be restructured to make the changes to the code and model straightforward in terms of a patch or a few additional commits.

[juniperlsimonis]

that makes sense. the thing we'll definitely need to keep an eye out for though is updated model output structure that updated code is pointed to (the timename element in the TS output in particular), as we'll need to update model objects wherever they live. i've updated the model objects in the v0.2.0 branch, so the ones you are calling up remotely here can get updated from those ones, yeah?

do note that we'll need to update some of the code and model objects with the jump to v0.2.0 (ref: #131 ), so i'm not sure which we want to do first?

I think this PR is independent enough (it only changes 2 files) that it can be rebased on top of 0.2.0; and the vignette should be restructured to make the changes to the code and model straightforward in terms of a patch or a few additional commits.