316 Introduction to Robust Estimation and Hypothesis Testing The function returns the various test statistics and corresponding critical values. The value of the test statistic, Q, for main effects for factor A, is returned in t3way$Qa, for factor B it is returned in t3way$Qb, and for factor C it is in t3way$Qc. The corresponding critical values are returned in t3way$Qa.crit, t3way$Qb.crit, and t3way$Qc.crit. The tests for two-way interactions are stored in t3way$Qab, t3way$Qac, and t3way$Qbc; the critical values are in t3way$Qab.crit, t3way$Qac.crit, and t3way$Qbc.crit; and the test for a three-way interaction is in t3way$Qabc, with the critical value in t3way$Qabc.crit. If data are stored in a matrix, with some of the columns indicating the levels of the factors, it is noted that the function fac2list, described in Section 7.1.2, can be used to store the data in the manner required here. Suppose the data are stored in a matrix, say m, with group numbers for the three factors stored in columns 2, 4, and 6. If, for example, a 2-by-4-by-5 design is being examined, column 2 would contain the group identification numbers for the two levels of the first factor. The values in column 2 might be 1 or 2, or they might be a 10 and 16. That is, there are two distinct values only, but they can be any two numbers. If the outcome measures are stored in column 5, the R command dat=fac2list(m[,5],m[,c(2,4,6)]) will store the data in dat, in list mode. If, for example, it is desired to compare 20% trimmed