Should I use relative risk ratio or marginal effects to interpret the output of multinomial logit?

I want to know which one is the best for interpretation of logit regression.

30 December 2014 2,783 5 View