NONPARAMETRIC MULTIPLE COMPARISONS OF SURVIVORSHIP DISTRIBUTIONS

Several common, nonparametric methods of analysis exist for determinin g significance among survival distributions of K groups or populations . One problem with these methods is that they cannot determine which s urvival distributions are different from one another if the global hyp othesis of equality is rejected (that is, if all distributions are not equivalent). We discuss the answer to this problem and illustrate the use of a little-known, nonparametric approach for making multiple com parisons among K > 2 survival distributions. This procedure can be use d with survival data that include censored observations. An observatio n on the survival time of an individual insect is censored if the stud y is concluded before death occurs, or if the insect is lost (for exam ple, escaped) for reasons not related to the experiment. The use of th e Sidak inequality to adjust the comparison-wise error rate is discuss ed. The procedure is illustrated using 2 entomological examples.