1.There are several alternative methods for calculating pseudo-R2 values in GLMs. One such approach in a dichotomous probit model is to construct a two-by-two table of observed and predicted outcomes, where individuals with Xβ > 0 assigned to have a predicted score of “1” (otherwise they receive a predicted score of “0”). The R2 is then simply the proportion of correctly classified individuals. Under a Bayesian approach, we would obtain multiple values for β and, hence, multiple possible R2 values. Perform this process and compare the result with what is obtained using the method I described in the chapter. How does the distribution of the outcome variable influence the difference between these two types of R2s?
2.Develop a strategy for handling missing data in the probit model (dichotomous or ordinal). Assume the data are MAR.