Table 3, Model 1 reports results of
regression models used for hypothesis
testing in Study 1. Because the dependent
variable was a count measure that
exhibited overdispersion, I estimated negative
binomial regression models. Hypothesis
1 posits a positive main effect of loss.
Because the model includes both loss and
loss 3 high rank, I tested and found support
for Hypothesis 1 by conducting a joint
Wald test of the significance of the two
covariates (p .001). Participants in the
gain conditions are predicted to choose 3.6
network interaction partners, while those
in the loss condition are predicted to discuss
the situation with 4.7 network interaction
partners. Consistent with Hypothesis 2,