variable selection procedure for a multivariable logistic regression model
Posted in the Epidemiology Forum
Since: Mar 13
#1 Mar 21, 2013
Here is the case,and I need some input:
I am assessing a bunch of risk factors and their associations with HIV infection(odds ratio will be the final measure).
Those factors include demographic characteristics, sexual behaviors and some other health related variables. I don't have a specific exposure in this casew; outcome is HIV infection(yes vs. No).
Normally, I will select a priori covariates and put them in the model based on DAG,biological mechanism or evidence from previously published journal articles, if I have a specific exposure and outcome.Then I will use backward selection method to retain those singnificant ones(based on 10% change-in-estimate rule of thumb).Apparently I don't think I can do it now because I don't have a specific exposure. What I am trying to do is to perform bivariate analysis of each factor with the outcome and pick those with p-value less than 0.1 to be included in the multivariable model.Then I will use backward selection procedure to generate a parsimonious model as the final model.However, this method is considered data driven by some epi people.
What do you think would be the better method for variable selection in this case?
Thanks in advance.
#2 Apr 12, 2013
first maybe you should do the regression plot with the Pearson value and after the reg log.
Add your comments below
|MMR coverage vs. confirmed cases||May 23||Sterling||2|
|Which epidemiological study design to choose?||Jan '17||sharu043||1|
|Economics of Epidemiology||Jan '17||maahrose||1|
|Need suggestions||Jan '17||maahrose||2|
|Epi Help (May '15)||Jan '17||mrose||2|
|Type of study??? (Jun '16)||Jan '17||Sajid Hasni||2|
|Measurement of Association (Sep '16)||Sep '16||PrettyErin||1|
Find what you want!
Search Epidemiology Forum Now
Copyright © 2017 Topix LLC