348 Introduction to Robust Estimation and Hypothesis Testing The command linconm(w,bend=1.1) would use an M-estimator, but the default value for the bending constant in Huber's would be replaced by 1.1. The function pbmcp(x, alpha = 0.05, nboot = NA, grp = NA, est = mom, con = 0,bhop = F, ...) performs multiple comparisons using method SR described in the previous section. (Method SR should not be used when comparing trimmed means.) By default, all pairwise comparisons are performed, but a collection of linear contrasts can be specified via the argument con which is used as illustrated in Section 7.4.1. With bhop=F, method SR is used, and setting bhop=T, the Benjamini­Hochberg method is applied instead, which is described in Section 7.4.7. 7.6.4 M-Estimators and the Random Effects Model Little has been done to generalize the usual random effects model to M-estimators. The approach based on the Winsorized expected value does not readily extend to M-estimators unless restrictive assumptions are made. Bansal and Bhandry (1994) consider M-estimation of the intraclass correlation coefficient, but they assume sampling is from an elliptical and