I’m asking this because I know there are some people here very knowledgable about maths and stats. 🙂
I’m running a logistic regression with a binary outcome and several categorical predictors of which one (gender) is my main interest.
Group A (male) has an outcome which is 5% yes and Group B (female) has an outcome which is 14% yes. This a risk ratio of 2.8 for B in reference to A.
If I run that predictor alone through the regression the Nagelkerke pseudo r2 says it only accounts for 5.4% of the variance.
Intuitively I’d feel that gender should account for more than that given the gulf in the outcome. I’m not formally trained in stats and somewhat floundering around here.
I have a handful of other predictors which when all put into the model still only accounts for 10% of the variance. Is there anything I can conclude about this? Model is useless? Gender isn’t much of a factor?