switch to using sklearn's pca

[ijyliu]

it seems to do better and we should do this for consistency

Do this in both

• [x] Simualtions code jupyter notebook to run locally
• [ ] Empirical applications code

There should be an example in my parallel run sims code

[ijyliu]

Ugh for the empirical code this could be a pain because we were using the built-in functions for the scree plot, etc from that other pca package...

[ijyliu]

So it's probably not worth it