st: RE: Automatic distribution fitting

[Nick Cox <n.j.cox@durham.ac.uk>]

I think the short answer for Stata is No in terms of an existing super-command. 

A longer, more personal answer: 

This sounds like a search for snake oil to me! That is, 

1. There are hundreds, thousands, ... of named distribution families. On the other hand, it may be that you can get useful results even from a program that looks at between (say) 10 and 100 families. 

2. There have been a few attempts to approximate many of them by superfamilies with several parameters but such procedures often seem to raise far problems than they solve. Google and then look at the various works of Karian and Dudewicz, for example. 

3. It would be a lot of work to replicate decisions that are easy for the experienced researcher, e.g. I only care about continuous/discrete, positive/nonnegative/..., etc. 

4. I'd be wary about using numerical criteria alone, even for distributions aimed at the same support with the same number of parameters. 

On the other hand, I don't know anything about EasyFit and whether it is smart enough to be very smart about these and other issues. 

Nick 
n.j.cox@durham.ac.uk 

Jorge Eduardo Pérez Pérez

Is there a command to do automatic distribution fitting in Stata ? I
know there are commands to adjust a given distribution to the data
(i.e -betafit- , -gammafit-) or to evaluate the fit of a given
distribution (i.e. -dpplot-, -hangroot-), but I would like to search
for the best fitting distribution over a set of theoretical
distributions, such as Mathwave's software EasyFit does.

To do this task, it is possible to write a wrapper command to the
distribution fitting commands already available. Has anyone already
done this?

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/