The impact of ignoring the stratification effect on the probability of
a Type I error is investigated. The evaluation is in a clinical setti
ng where the treatments may have different response rates among the st
rata. Deviation from the nominal probability of a Type I error, alpha,
depends on the stratification imbalance and the heterogeneity in the
response rates; it appears that the latter has a larger impact. The pr
obability of a Type I error is depicted for cases in which the heterog
eneity in the response rate is present but there is no stratification
imbalance. Three-dimensional graphs are used to demonstrate the simult
aneous impact of heterogeneity in response rates and of stratification
imbalance.