The hypothesis test for two-way ANOVA has three null hypotheses, each of which can be retained or rejected. The decision rule must mention the decision criteria and critical value for rejecting each hypothesis, and each must be mentioned in the decision and interpretation. For a review, see the hypothesis test summary in your text. The variance estimates in two-way AN OVA are similar to those in one-way ANOVA in that measures of variability always consists of a variance estimate of (14) calculated by dividing the sum of squares by its (15) . As in oneway AN OVA, the major computational effort is in calculating the various (16) . For computational formulas for the various SS terms and df terms, see Table 23.2. The same procedures used for one-way ANOVA also apply to two~way for obtaining critical F values from Table C, Appendix D. Whenever interaction is present, as indicated by the (17) of the null hypothesis for interaction, special statistical tests are usually conducted to pinpoint the precise location of the discrepancies that caused it. These special tests may be explored further in the Kepple book footnoted in your text. Differences between (18) means and column means may also be examined with a modication of Sceffe's test for multiple comparisons. 29