[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: Re: Two postings to SSC

Subject   st: Re: Two postings to SSC
Date   Mon, 19 Oct 2009 11:34:55 -0400

I should like to make you aware of two statistical commands that Kit was kind enough to place on the SSC website for me. One was placed several weeks ago, the other yesterday.
 TPOISSON - truncated Poisson regression
Program allows one to left or right truncate the response of a Poisson model, or truncation can be made to both sides in the same model. Full Stata maximum likelihood options are available; eg survey,
      bootstrap, etc

 CPOISSONE - censored Poisson (econometric parameterization)
Program allows both left and right censoring of a Poisson model, or censoring can be made to both sides in the same model. I differentiated between two types of censored Poisson models in my book, "Negative Binomial Regression" (2007, Cambridge University Press).

In the mid 1990s I wrote a censored Poisson as a SAS macro (still published on the SAS Library site), then a year later Dean Judson and I wrote a censored Poisson for the Stata Technical Bulletin. I later revised it, and did so again in 2005. That version of a censored Poisson let any observation value be censored, and was parameterized as censoring occurs in survival models. This differs from the parameterization that has been common in econometric literature, where censoring occurs only at the tails of the response distribution. That is, left censoring, like left truncation, can occur at time=3, with all observations below 3 being censored. Likewise for right censoring; all values over a specified cut point are censored. It is similar to truncation, except that it differs on how censored observations are valued. In econometric censoring, censored values are recorded as the value of the cut point. In truncation, the truncated values are dropped from contributing to the model, but the LL function is
      adjusted to still sum to 1.

The previous survival parameterization of censored Poisson is on the SSC site as CPOISSON. The econometric parameterization of censored Poisson is CPOISSONE, and truncated Poisson is TPOISSON. I am planning to prepare econometric parameterized censored negative binomial and truncated negative binomial commands next. I already have a survival censored negative binomial on SSC called CENSORNB, posted in 2005.

Joseph Hilbe

*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index