Which statement best defines the Tukey-Kramer test?

A nonparametric rank-based test used to compare medians of many groups after ANOVA indicates unequal variances and normality assumptions fail.

A method for adjusting pairwise p-values using a pooled variance estimate but applied before conducting ANOVA to decide if an ANOVA is necessary.

A post hoc technique that sequentially removes the smallest group means and then performs t-tests between the remaining groups, requiring adjustments on each step to control overall Type I error across the sequence of removals, especially when group sizes differ and heteroscedasticity is suspected — used after a significant ANOVA.

A post hoc multiple-comparisons procedure applied after rejecting an ANOVA null that tests all pairwise mean differences while controlling the familywise error rate.