Odds ratio of a continuous covariate

questions concerning analysis/theory using program PRESENCE

Odds ratio of a continuous covariate

Postby pennyb » Mon Nov 02, 2015 5:40 am

Hi all,

I've been working through how to determine the effect of a covariate on psi, however have only come across worked examples that use categorical covariates (e.g. browsed/unbrowsed for the Weta example and good/poor colony for the Swift example in MARK). I understand it thus far that to calculate an odds ratio, you are comparing relative successes and failures of two 'outcomes', but what if you have a continuous covariate, such as habitat patch size?

I am trying to figure out the 'effect' (if any) of patch size (standardised into a z score) in my data, but haven't grasped how to apply the odds ratio principle to this kind of data.

Any thoughts would be much appreciated.

Cheers, Penny.
pennyb
 
Posts: 13
Joined: Tue Sep 23, 2014 9:06 pm

Re: Odds ratio of a continuous covariate

Postby Alan Dextrase » Mon Nov 02, 2015 2:45 pm

Hello Penny,

For a continuous covariate, the covariate's coefficient represents the change in log odds for a 1 unit increase in the covariate. Since you are using a standardized covariate, it would be the change in log odds as the covariate increases by an amount equivalent to the covariate's standard deviation.

Alan
Alan Dextrase
 
Posts: 17
Joined: Tue Dec 07, 2010 10:01 pm
Location: Peterborough ON, Canada

Re: Odds ratio of a continuous covariate

Postby pennyb » Tue Nov 03, 2015 1:21 am

Thank you Alan!! :D
pennyb
 
Posts: 13
Joined: Tue Sep 23, 2014 9:06 pm

Re: Odds ratio of a continuous covariate

Postby pennyb » Tue Nov 03, 2015 2:34 am

Alan Dextrase wrote:Hello Penny,

For a continuous covariate, the covariate's coefficient represents the change in log odds for a 1 unit increase in the covariate. Since you are using a standardized covariate, it would be the change in log odds as the covariate increases by an amount equivalent to the covariate's standard deviation.

Alan



1 step forwards..2 steps back..I have just come across another related but slightly different query...

If I want to look at an effect of a categorical variable that has more than one parameter (e.g. say 3 habitat types), I gather that to get the odds ratio you would calculate exp(B0)exp(B1)exp(B2), however, how would a confidence interval be calculated if each beta has its own standard error?
pennyb
 
Posts: 13
Joined: Tue Sep 23, 2014 9:06 pm


Return to analysis help

Who is online

Users browsing this forum: No registered users and 9 guests