[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
st: -pgmhaz- updated to version 8.2
I have updated my -pgmhaz- program to version 8.2. The new version is
called -pgmhaz8- and, thanks to Kit Baum, may be installed from SSC
using the command -ssc install pgmhaz8-. (Users of Stata versions 5.0
to 8.1 may continue to use -pgmhaz-; it can be installed using -net stb
-pgmhaz8- estimates by ML two discrete time (grouped duration data)
proportional hazards regression models: (1) the Prentice-Gloeckler
(1978) model; and (2) the Prentice-Gloeckler (1978) model incorporating
a Gamma mixture distribution to summarize unobserved individual
heterogeneity ('frailty'), as proposed by Meyer (Econometrica 1990).
Covariates may be fixed or time-varying; and the baseline hazard
function may be parametric or non-parametric.
-pgmhaz8- takes advantage of new features added to -ml-: for example,
users may use maximization algorithms other than the default NR one
using the -technique()- option, and convergence should be somewhat
faster than before. The program now reports the correct p-value for the
LR test of Model 1 versus Model 2. You can choose not to display Model 1
results, and fweights and iweights are allowed.
The program still does not allow robust or cluster() options. These, and
further reductions in convergence time, are not possible without a major
re-write -- converting the program from using numerical derivatives
(method d0) to using analytical derivatives (e.g. method d1). I started
on this, but ran out of time. (The expressions for the derivatives are
really complicated, so I may not have got them right anyway. If anyone
wants to compare their math with mine, let me know!)
As usual, comments and suggestions to me please.
Professor Stephen P. Jenkins <email@example.com>
Institute for Social and Economic Research
University of Essex, Colchester CO4 3SQ, U.K.
Tel: +44 1206 873374. Fax: +44 1206 873151.
* For searches and help try: