2. For binary data, define a generalized linear model with the log link. Discuss the effects refer to the relative risk. Why do you think this link is not often used? (Hint: What happens if the linear predictor takes a positive value?.)