Single-Stage multiple comparison procedures under heteroscedasticity

Given k (≥ 2) independent normal populations with unknown means and unknown (and possibly unequal) variances a single-stage sampling procedure for multiple comparisons with the largest normal mean and with a control, respectively, are proposed. The advantage of the single-stage procedure is to have design simplicity and to reach an exact solution. Simulation results indicate that the single-stage procedure appears to be a reasonable choice as compared to existing methods. Computer program and a numerical example are given.