Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

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

Re: st: log normal and logistic hazard ratio

From   Maarten Buis <>
Subject   Re: st: log normal and logistic hazard ratio
Date   Mon, 20 Jun 2011 08:37:44 +0200

On Mon, Jun 20, 2011 at 4:42 AM, Yusvita Triwiadhian S. wrote:
> Lognormal and logistic distribution for survival anlysis,there isn't
> hazard ratio,is it right? why? i'm looking for command to get its
> hazard ratio but i do not get it..

Because a hazard ratio are only reported in models that make the
proportional hazard assumption, and both the lognormal and the
loglogistic only work with the accelerated failure time metric.

In models that make the proportional hazard assumption you have a
baseline hazard function that says how high the hazard is at each
point in time for someone that has all zeros on his/her explanatory
variables. The proportional hazard assumption implies that the hazard
function of all other groups only differ from the baseline hazard
function by a fixed proportion. If we did not make that assumption
than our effect of a variable would differ depending on how much time
the individual has been at risk. It is thus because of this assumption
that we can summarize the effect in one number. Alternatively, if we
want to relax the proportional hazard assumption we can add
interaction terms with analysis time and thus let the effect change
with time, as I proposed here:

In accelerated failure time models like the lognormal and loglogistic
the idea is that times goes faster for some types of individuals than
others (e.g. one year for a dog is equivalent for 7 years for humans).
Again there is a baseline function, but now the explanatory variables
influence the time axis instead of the hazard.

Hope this helps,

Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index