1. Section 2.2.7 discusses the criteria for how R determines if and when the convergence of parameter estimates takes place. Discuss why the residual deviance is used for this purpose. Suggest other ways to determine convergence.
2. What does the na.action argument of the glm() function control and what is its default? Discuss possible problems that may occur with this default value.
3. Consider the complementary log-log regression model of log[-log(1-π)] = β0 +β1x1. Show that the odds ratio comparing different levels of x1 cannot be written without the explanatory variable being presented.