Move to static branching

Motivation-and-Behaviour / sleepIPD_analysis

Analysis for the sleep and physical activity pooled study (https://osf.io/gzj9w/)

Other

0 stars 0 forks source link

Move to static branching #87

Closed tarensanders closed 1 year ago

tarensanders commented 1 year ago

Switch to a static branching model for models/figures/tables
Add log pa volume as the default
Add sensitivity for scale pa volume

Remaining few jobs:

[x] Check if there are other bits that should be added from #85 (I've been adding @conig as co-author for these bits)
[x] Seem to be some differences in the figure filenames that have broken the multiverse code.

tarensanders commented 1 year ago

I think I've moved all the relevant bits from #85 over, but let me know if I missed anything.

The static branching uses the models_df tibble that's defined in the _targets.R file. We could move this to a different file. I do like how explicit it makes the definition of the models.

Also, it might be worth running tar_destroy() and starting from scratch on your next run. I spent ages trying to work out why some sections of the multiverse weren't getting updated, and it was because tar_load(target_thats_been_renamed) works. You probably want to get rid of any old targets.

conig commented 1 year ago

I don't know about you, but the data imputation step is now much quicker with that fix for my computer. I'm getting that target done in 3 mins now which is a relief. These are great changes, although I can feel your pain, this was a ton of work!

I agree the tribble makes it much more obvious what's happening with the target generation. I have a preference for it being in another file as it's kind of chunky, but it doesn't really matter. Check out the measurement day comment above but otherwise this is ready to merge from my perspective.

tarensanders commented 1 year ago

I don't know about you, but the data imputation step is now much quicker with that fix for my computer. I'm getting that target done in 3 mins now which is a relief.

Yeah, I messed up and each core was doing three imps which was why there was nine in total. Fixing that cuts 2/3rds of the time out haha.