Contributing an Example for GMM 2SLS Regression

[Vincent La]

Hi,

I was poking around with the sandbox gmm IV2SLS method, and I found the documentation to be a bit sparse and the examples to be a bit lacking. I think I finally did figure it out, and I'd be happy to write up an ipython notebook with an example and a bit of the stats background. http://www.ats.ucla.edu/stat/stata/examples/methods_matter/chapter10/default.htm seems to be the canonical example given in Stata, I could piggy back off that.

Thoughts? I'd like to contribute even if it is just a small thing!

Vicent

[josef...@gmail.com]

I don't know that textbook, but any improvements to the documentation and additional notebooks would be great, especially given the current sparseness.

Depending on license compatibility we can either add notebooks to the documentation or link to them in a separate repsitory or gist.

I'm ready to help if you have questions, or some things to look at. 

Also, based on a quick skimming of the UCLA page, we should add some information about the first-stage regressions, which we currently don't produce because we use simpler linear algebra. We could redo some of the first stage results using OLS for optional diagnostics. 

Thanks,

Josef

 

Vicent

[josef...@gmail.com]

On Mon, Nov 7, 2016 at 1:37 AM, <josef...@gmail.com> wrote:

On Mon, Nov 7, 2016 at 1:21 AM, Vincent La <vinc...@gmail.com> wrote:

Hi,

I was poking around with the sandbox gmm IV2SLS method, and I found the documentation to be a bit sparse and the examples to be a bit lacking. I think I finally did figure it out, and I'd be happy to write up an ipython notebook with an example and a bit of the stats background. http://www.ats.ucla.edu/stat/stata/examples/methods_matter/chapter10/default.htm seems to be the canonical example given in Stata, I could piggy back off that.

Thoughts? I'd like to contribute even if it is just a small thing!

I don't know that textbook, but any improvements to the documentation and additional notebooks would be great, especially given the current sparseness.

a general tip: the unit tests also contain useful examples for many models, but not always with real data.

Josef