Hi Paaveen,
I read your mail and have the following suggestions to offer.
Do stepwise selection procedure in logistic regression with all potential covariates. The procedure will eliminate any variable that is linearly correlated with others already in the model.
Your final variables selected through the stepwise selection criterion will be viable predictors without any collinearity. When doing stepwise selection, use alpha that is more conservative so that you have worst scenario included in variable selection.
Remember that your descriptive Statistics is just a stepping stone and not a conclusive stage of determining which variable goes into the model and which does not.
Mubashir
From: A UK-based worldwide e-mail broadcast system mailing list [mailto:[log in to unmask]] On Behalf Of paaveen jeyaganth
Sent: Wednesday, May 25, 2016 10:31 AM
To: [log in to unmask]
Subject: logistic regression
Dear Allstat,
I am doing logistic regression
before i do logistic regression. i did some descriptive statistic
death is outcome
N=450
covariate :- age (continuous), gender (2 group), HTN(2 group), PDM(2 group),
PMI(2 group) , ACS(3 group), bif(4 group), GEN(2 group)
for descriptive statistics , i did chi-square test GEN is related to covariate
from that i found ACS and HTN are statistically significant to GEN,
when i do the backward logistic regression
1) is it correct include ACS , HTN and GEN in the model
or do i have to only include one of these above variable?
2)When i do the logistic regression i include all the variable in the
model final model include ACS and GEN
if this two variable are related to each others then there is collinearity is present?
i really need step-wise advice what i am doing correct?
appreciate any advise,
thanks
paaveen
You may leave the list at any time by sending the command
SIGNOFF allstat
to [log in to unmask], leaving the subject line blank.
SIGNOFF allstat
to [log in to unmask], leaving the subject line blank.