Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

RE: st: Adjusting for age in logistic regression

From   Leonelo Bautista <[email protected]>
To   [email protected]
Subject   RE: st: Adjusting for age in logistic regression
Date   Thu, 11 Nov 2004 09:00:57 -0600

Including age in your model will adjust for that variable. Now, Maarten
points to an issue that goes beyond what would be the reasonable age of
subjects included in your study. I mean, you should check whether age is
linearly related to the log odds of disease before including it in your
model as a continuous variable. You can do this easily by using -lintrend-.
If age is not linearly related to the log odds of disease, then one option
would be to categorize age and use dummy variables. Also, depending on what
you are studying and your sample size, it may make sense to adjust by age as
a categorical variable.

Maarten argues that after including age in your model, you can still have
"problems". I reckon he is referring to residual confounding (confounding by
unmeasured or poorly measured variables). Since there will always be
unmeasured potential confounders and there will always be some error in our
measurements (at list in the context of epidemiologic studies), residual
confounding is something one can argue in practically any situation. The key
point here is whether residual confounding is likely to be so large as to
invalidate your conclusions. 

Leonelo E. Bautista
University of Wisconsin Medical School
Population Health Sciences

-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of maartenbuis
Sent: Thursday, November 11, 2004 4:16 AM
To: [email protected]
Subject: Re: st: Adjusting for age in logistic regression


There are two issues here: First, should age be added as a single 
linear variables. Lets say you are interested in whether or not a 
women gets a child, and the variable age has a range of 0-100 years. 
Adding it as a linear term would clearly be nonsense. In this case I 
would start with creating a series of dummies: say 0-12, 12-18, 18-
30, >30. So how you want to include a variable depends on how you 
think the variable effects the probability of experiencing the event 
(= your theory). 

Second, some recent posts have suggested that there are still 
problems after you correctly entered the controll variable: see


> I am doing a logistic regression to determine the effect of risk 
> factors to the outcome. What I want to do is adjust for age and I 
> am including the age variable (continuous) into the model. What I 
> am not sure of is if this is the right approach.

*   For searches and help try:

*   For searches and help try:

© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index