Bayesian PCA

[Al Bothwell]

Hi Everyone!

I am wondering if bayesian principal components analysis (PCA) is possible in Stata.

Al Bothwell

[Nick Cox]

My prior is that there is no existing code if search or Google does not find it.

[Jared Greathouse]

My posterior is that Stata is still pretty limited in terms of its ML/ data science capabilities. Python, R... those likely have what you seek, which you can do from Stata directly, if you know python.......

I only was able to program a PCR regression estimator (a form of synthetic control estimation) earlier today BECAUSE I was able to see the raw python code and how it worked enough to adapt it to a causal inference setting. So, if you have Python, I'm pretty sure they'd have it, if it exists. Then you can just use Python from your do file and boom. Either that, or get into all the scary Mata you must have to know in order to even try it or attempt it.....

But as for native Stata, or user written commands? I'd wait until Stata 20.


EDIT: Python has what you seek (here https://github.com/Duuuuuu/Probabili...d-Bayesian-PCA)

But I have no idea how you'd even use this, or for that matter why you'd want to. I'm frankly not smart enough or mathematically gifted to understand it, but just know the option is on the table if you'd wish.

[Al Bothwell]

Thanks for the comments. Glad to notice that we can have stats fun re: "prior" and "posterior" in #2 and #3.