st: Stata 13

[Maurizio Pisati <maurizio.pisati@unimib.it>]

Dear Rafal,
here are some of my ideas for Stata 13.
Best,
Maurizio

IDEAS FOR STATA 13

1. New commands
- Latent class analysis
- Multilevel multinomial logistic regression (-xtmemlogit-)
- Multilevel generalized linear models (-xtmeglm-)
- Commands for fitting distributions regression-wise (see, e.g., user-written commands -skewnreg-, -skewtreg-, -invgammafit-, -betafit)


2. Improvement of existing commands
- Extend the logic of factor variables to quantitative variables, introducing operators that represent various kinds of transformation (e.g.: cs#.varname -- cubic spline with # nodes; ln.varname -- natural logarithm). This would make it easy to use -margins- to get quantities of interest for quantitative variables (with the current set of operators, only polynomials are handled nicely by -margins-).


3. Documentation
- A "Mata for dummies" manual -- unveiling Mata's potentialities in a plain language and with a lot of practical examples (Bill's way, so to speak) -- would be a great asset for many Stata/Mata programmers.