TWIST on Cox Regression for Conditional Logistic Regression - Need Help

[cmuegge]

I have a unique use of CLR  - my study compares excess mortality in firefighters versus non-firefighters.  I matched FF to nonFF on age, sex, race, ethnicity and year of death.

I want to know if occupation is related to cause of death.  The outcome is cause of death but not all FF have the outcome, not do all NonFF.  My dissertation committee tells me to CLR... I think I have it, but not sure.  Any help you might give me would be totally awesome!

___

[Bruce Weaver]

Do you mean that your committee is advising a *preliminary* CLR with death as the outcome BEFORE you proceed to your main analysis, which has cause of death as the outcome variable?  That's the only way I can make sense of what you've written.  

If that is what you mean, the link below shows the old work-around of using COXREG to perform conditional logistic regression.

http://spsstools.net/en/syntax/syntax-index/regression-repeated-measures/conditional-logistic-regression/

But another option, that is arguably more straightforward, is to use GENLIN with generalized estimating equations (GEE) to account for the correlated nature of the data.  In some analyses I did several years ago, this approach yielded results very similar to CLR performed via Stata.  

http://spssx-discussion.1045642.n5.nabble.com/Conditional-logistic-regression-td5729377.html

Finally, how many categories are there for your "cause of death" variable?  And what does a cross-tab of profession by cause of death look like?  Are all of the cells reasonably well populated?

HTH.

cmuegge wrote

I have a unique use of CLR  - my study compares excess mortality in firefighters versus non-firefighters.  I matched FF to nonFF on age, sex, race, ethnicity and year of death.

I want to know if occupation is related to cause of death.  The outcome is cause of death but not all FF have the outcome, not do all NonFF.  My dissertation committee tells me to CLR... I think I have it, but not sure.  Any help you might give me would be totally awesome!

___