The two-step approach—a significant ANOVA F-test before Dunnett's comparisons against a control—is not recommended