Add functionality for Frequentist IV regression classes

[NathanielF]

Adding the Frequentist class for implementing IV regression experiments as discussed here: #213

[NathanielF]

Is this a good place @drbenvincent , @juanitorduz to discuss the addition of statsmodels to the CausalPy backend? Or should we pull it out into it's own issue?

The case for it here is largely that it would help with the statistical output of IV treatment effects and potentially combines really nicely with stargazer. To present multiple regression model summary tables that allow you to easily inspect the addition of different covariates to your regression model.

[juanitorduz]

I think linearmodels is a fantastic package for that use case. WDYT?

[drbenvincent]

There is already an issue for statsmodels backend so we can use that. Or a new issue for a linearmodels backend.

Hugo is 7 days old today, so forgive any slowness in responding on my side. We're still near 100% baby logistics.

[NathanielF]

Wasn't actually expecting a quick reply @drbenvincent !! Just putting a note here for discussion.

I don't really know the linearmodels package well but happy to consider it. Will pick this up in the existing statsmodels issue.

Can explore both options on this issue.

[NathanielF]

Coming back to this one. I've had a look at linearmodels as a package. It's great. Has various different ways that you can implement frequentist IV regressions. It has neater and clearer than the documentation than that available for statsmodels. I think i would like to implement the frequentist IV class with linearmodels.

https://bashtage.github.io/linearmodels/iv/examples/advanced-examples.html

How does/should this impact the discussion of the desired backend for frequentist classes?