Observations from the ith treatment and the jth block are randomly generated from N(i + j, σ 2)
in order to simulate designs with different treatment effect size and block effect size. When the null
hypothesis is true, we set σ2 = 1. The variances under the alterative hypothesis are listed in the
second column of Tables 2 and 3. We set the nominal null rejection rate to be 0.05 and repeat this
procedure for 10,000 times to determine the simulated rejection rates of the null hypothesis. The
results are listed in Tables 2 and 3. In each scenario, the first row shows the Type I error simulation and
the remaining rows are for power simulation. The best test with controlled Type I error and the highest
power among 9 tests were marked in red. The number in blue indicates that the corresponding test
was unable to control Type I error near or under nominal rate α = 0.05. As a result, the subsequent
simulated powers were marked in grey and not considered for power comparison among different
methods.