The type I error rate of the new Fmax- test is well controlled under different block size and treatment
sizes. However, Levene’s test, Han’s Fmax test and Yitnosumarto’s |z| and z2tests were not able tocontrol Type I error rate near or under the nominal rate 0.05 when the treatment size is large and the block size is relatively small. As shown in Table 3, the null rejection rate under design scenario III for Levene’s test, Han’s Fmax-test, Yitnosumarto’s |z| and z2tests were greater than 0.1. As blocksize (within treatment sample size) increased to 10 in Scenario IV, the null rejection rate under designscenario III for Levene’s test, Han’s Fmax test, and Yitnosumarto’s |z| and z2tests got better but werestill greater than 0.07. As block size (within treatment sample size) increased to 20 in Scenario IV, the