Muddy Points

Lesson 15: Other regressions

Muddy Points from Spring 2025

With multinomial regression so we are comparing several groups to each other depending on the variable and number of categories is this correct?

My muddy point is why logistic regression gives us log odds/odds ratio (exp(B_k)) and log-binomial gives us risk ratio. Does it simply have to do with the fact we are using logit(µ) and log(µ)?

I need more practice on figuring out when log-binomial and logistic regression are preferred over each other – recalling that we can transform risk ratios to odds ratios…

Still a little unclear of why you would use the log link instead of the logit link for the Bernoulli/Binomial distribution