Effects of Helper and Recipient Sex 19
The hypotheses were addressed initially through two multivariate analyses of variance
(MANOVAs): one analysis on the two primary outcome variables that assessed recipient affect
states (affective improvement and perceived comforting quality of the helper’s behavior), and the
second analysis on the three assessments of helper communicative competence (perceived
sensitivity, conversational engagement, and normativeness). Both sets of analyses reflected a 2 x
2 x 3 x 3 (recipient sex x helper sex x VPC x NVI) design. In cases where significant
multivariate effects were observed for the factors, Roy-Bargmann stepdown tests were conducted
to control for redundancy among the dependent variables. Statistical power (with = .05) for the
main effects of recipient and helper sex was .99 for large effect sizes (f = .40) and .93 for
medium effect sizes (f = .25). Power for the main effects of VPC and NVI was .99 for large
effect sizes and .86 for medium effect sizes. For the theoretically important two-way and three-
way interactions that involved sex of recipients and/or helpers and message form (either VPC or
NVI), power was also .99 for large effect sizes and .86 for medium effect sizes. For the two-way
interaction between the two message factors, the three-way interactions that involved both
message factors, and the four-way interaction among all the factors, power was .98 for large
effects and .78 for medium effects.
Analyses for Affective Improvement and Comforting Quality
The MANOVA on affective improvement and perceived comforting quality detected a
significant multivariate main effect for VPC, Wilks’
= 0.82, F (4, 358) = 9.14, p < .001,
2
=
.09, as well as a significant main effect for NVI, Wilks’
= 0.94, F (4, 358) = 2.96, p < .02,
2
= .03. In addition, the multivariate main effect for recipient sex approached significance, Wilks’
= 0.97, F (2, 179) = 2.55, p < .09,
2
= .03. No other multivariate effects approached
Convention