Tools for controlling Stata from Python. Reads log files, generates dynamically customizable tables (in cpblTables format) from regression output; other tools
There are three repositories
pystata cpbl-tables cpblUtilities
now populated on my github account (cpbl) which, along with a file cpblDefaults.py (a version of which is in pystata) make a minimal set of stuff to run the pystata-demo.py program in pystata repository.
As you will see, making this code open is a bit bold because it's a decade of personalized kludges (messy), with some useful tools interspersed. :)
Here's the idea of one component of pystata, the latexRegressionClass:
I write regression commands like in a Stata do-file, but with some extra annotation. E.g. lines like the following could preceed a regression call
name:Full model flag:state dummies *flag:cluster=household
and these give some control over annotation in output tables.
Preprocessing of the stata commands by my class adds numerous extra features to the eventual Stata log file (such as eliciting the covariance table). Stata code in multiline text format such as the above is converted into a models data structure in python, and the estimate results appear in this structure when they're available.
Sets of estimates are grouped together into tables, and the stata code collected by the class. At some point, this code is executed as a batch Stata job.
The model estimates are available for plotting or further analysis in python, including in Pandas format, and are also used to generate LaTeX tables of results.
There is much customization of these tables possible, including substituting Stata variable names for LaTeX markup, specification of ordering of variables, and so on, which can be refined without waiting for the estimates to be rerun.
The LaTeX tables are stored individually, each in a custom format .tex file which actually includes, as comments for posterity, the entire Stata log file lines for the relevant estimates.
These .tex files are incorporated into LaTeX articles, beamer presentations, or etc, using the cpbl-tables style files, and they allow on-the-fly reformatting into various kinds of tables, such as wide, long, landscape, etc, simply by changing the call function in your main document. By adding "transpose" to the name, you can even invoke a transposed version of the table.
That is the gist of what the pystata-demo.py is supposed to show, and of course there are many other things in each of the packages.
I'm sure you'll have lots of fiddling to even get stuff set up at first; I hope it's manageable and worthwhile.
The hope is to sort out the idiosyncratic from the useful/general, and the good from the old/bad.