Free Trial

Safari Books Online is a digital library providing on-demand subscription access to thousands of learning resources.

Share this Page URL

8.6 Between-by-Within and Within-by-With... > 8.6.16 Multiple Comparisons when Usi... - Pg. 429

Chapter 8 Comparing Multiple Dependent Groups 429 the other. For the data in Table 6.6, it was found that schizophrenics differed from the control group; also see Figure 6.10. If the two groups are compared based on the OP estimator (using the function smean2), again the two groups are found to differ. Comparing the groups with the method for means and trimmed means described in this section, no difference between the schizophrenics and control group is found at the 0.05 level. Using the rank-based method in this section, again no difference is found. (The p-value is .11.) The only point is that how we compare groups can make a practical difference about the conclusions reached. I 8.6.14 Rank-Based Multiple Comparisons Multiple comparisons based on the rank-based methods covered here can be performed using simple combinations of methods already considered. When dealing with Factor A, for example, one can simply compare level j to level j , ignoring the other levels. When comparing all pairs of groups, FWE can be controlled with Rom's method or the Benjamini­Hochberg technique. Factor B and the collection of all interactions (corresponding to any two rows and any two columns) can be handled in a similar manner.