Odds Ratio in Logistic Regression?

Buckeye

Active Member
#1
I will get to my question in a bit. But first, some introduction. The regression coefficients describe the change in log odds of the event for a one-unit change in the independent variable. In the case of a categorical predictor, the coefficients describe the change in log odds of the event for that level relative to the baseline level. And we can create nice dot and whisker plots like this: https://www.r-bloggers.com/2021/06/oddsplotty-has-landed-on-cran/

We know that odds = probability for event / (1-probability for event). And a CI that lies above 1 corresponds more favorably to the event of interest while a CI that lies below 1 corresponds more favorably to the other event. A CI that includes 1 suggests the predictor is not significant in predicting the outcome.

My question: is there a straightforward way to visualize the probabilities similar to the odds ratios? I feel that my audience will understand probabilities better than the odds ratios. I have multiple predictors so I don't think showing the logistic S-curve plot works in this case.
 

hlsmith

Less is more. Stay pure. Stay poor.
#2
Yes, you calcalute predicted probabilities with CI's. In your case these would be conditional.

I haven't used it but people love emeans in R. Maybe two M's in it.

This can be nice since they are bound and not 0-infinity.