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:ODDSRISK module and continous covariates

From   Ronan Conroy <>
To   "" <>
Subject   Re: st:ODDSRISK module and continous covariates
Date   Fri, 4 Feb 2011 12:05:36 +0000

On 4 Feb 2011, at 05:44, Nanlesta Pilgrim wrote:

> I'm  new to the ODDSRISK module.  My dataset is longitudinal and I'm
> using log binomial regression (specifically GEE) to model my outcome
> which is not rare.  The problem is that the adjusted model does not
> converge.  As such, we decided to do logistic regression and then
> convert the ORs to RRs using Zhang's formula.

The trouble with risk ratios is that they can end up predicting risks that go beyond the range of zero to one. While univariate risk ratios are very easy to communicate, they have to be taken with a pinch of salt. For example, smoking has a risk ratio of 2.5 for heart attack, but if your risk of a heart attack, based on your other risk factors, is 60%, then smoking increases this to 150%. 

Looking at your output, you have a lot of variables with very wide confidence intervals whose inclusion in the model seems to be of questionable value, since they are going to reduce the precision. 

Ronán Conroy
Associate Professor
Division of Population Health Sciences
Royal College of Surgeons in Ireland
Beaux Lane House
Dublin 2

*   For searches and help try:

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