'Duplicate the small simulation study reported in Table 20.2 on page 613, comparing several methods of handling univariate missing data that are MAR. Then repeat the study for missing data that are MCAR and for missing data that are MNAR (generated as in Figure 20.1 on page 608). What do you conclude? Note: This is not a conceptually difficult project, but it is potentially time-consuming; it also requires some programming skills and statistical software that can generate and analyze simulated data.