[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: predictions from ML program

From   "Stephen P. Jenkins" <>
To   <>
Subject   st: predictions from ML program
Date   Thu, 12 Jun 2008 15:57:24 +0100


Date: Wed, 11 Jun 2008 07:51:53 +0000
From: marcel spijkerman <>
Subject: RE: st: predictions from ML program

Predict is indeed useful but what I need is the contribution to the
likelihood for each of the four terms in the likelihood (in:

ln(p1*L1 + p2*L2 + p3*L3 + (1-p1-p2-p3)*L4)

The model is a bivariate survival model with discrete heterogeneity
distribution; and the terms in the likelihood are contributions for each
person of the two duration processes. Can I obtain the separate
contributions L1, L2, L3 and L4 by using predict or other postestimation

At the top of your ML evaluator program, you should have statements like

	mleval `IA' = `b', eq(1)
	mleval `IB' = `b', eq(2)

These are linear indices of the type discussed before, and which you want.

And later in the evaluator program, you will generate likelihood
contributions which refer to `IA' and `IB' (and other things like the
probability parameters and mass points)

I recommend that you do not go any further without consulting (a) some
existing programs using -ml- to see these ideas in practice, and especially
(b)  Gould, W., J. Pitblado, and W. Sribney. 2006.  Maximum Likelihood
Estimation with Stata. 3rd ed.  College  Station, TX: Stata Press. (Or
earlier editions.) Available from Stata Press website.

Professor Stephen P. Jenkins <>
Director, Institute for Social and Economic Research (ISER)
University of Essex, Colchester, CO4 3SQ, UK
Tel: +44 (0)1206 873374. Fax: +44 (0)1206 873151
Survival Analysis using Stata:

Learn about the UK's new household panel survey, the United Kingdom
Longitudinal Study:

*   For searches and help try:

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