|
Consider again the situation described in Exercise 11. This time, use simulation to assess the performance of the usual two-sample t test. That is, use the same simulations as in part (c) of Exercise 11 (or ones just like them if you do not have the same simulations). This time, for each simulated sample compute the statistic U in Eq. (9.6.3) and keep track of the proportion of simulations in which U is greater than each of the nominal t quantiles, T−119 (1− α0) for α0 = 0.1, 0.05, and 0.01. How do these proportions compare to the nominal α0 values? |
| New search. (Also 1294 free access solutions) |