Effect Size measure for Ordinal Logistic Regression

[LycianSage]

Hello friends,

I am using "Ordinal Logistic Regression" for my dissertation research. I wanted to report the effect size but couldn't find definitive reference for reporting an effect size.

I have found some papers stating that "Nagelkere's R square" is the effect size measure. Another states "odds ratio". Interestingly, I found other papers saying there is no effect size for logistic regression in general (ordinal, multinomial, etc)[1,2].

So, what is the effect size measure for "Ordinal Logistic Regression" and what is your reference? (yes, please cite the paper)

Thank you in advance for any help.

___
[1] Menard, S. (2000). Coefficients of determination for multiple logistic regression analysis
[2] Peng CYJ et al. (2002). An introduction to logistic regression analysis and reporting

[LycianSage]

any idea? come on...

[bryangoodrich]

Which effect size you want to report depends on a number of things. For instance, I work with an epidemiologist, so the results tend to be on health topics or mortality. As such, reporting effect size as risk ratios tends to send a clear message about the prevalence of such and such increasing mortality or health risks. One can conceivably think there are other ways to report effect sizes from a logit model, but that is contextual and depends on the outcome and your audience.

[LycianSage]

I'm studying the association between personality traits and the use of an internet application.

My sample is comprised of undergraduate university students. Personality traits are measured by NEO-FFI personality scale. It has 5 dimensions which I extracted by Factor Analysis. Latent personality variables (IV) are constructed by the "regression" method of factor analysis by SPSS. "use of the software" variables (DV) are various. there are continuous, nominal and ordinal ones. I will use "Ordinal Logistic Regression" for the ordinal DVs.

I would be greatful for any help you can provide. I can't find someone who knows OLR...

[Kasperat]

Did you ever solve this question? Which effect test did you use and why?

I am struggling with a similar challange :-)