In the second step of the model, we examined the effect of smile intensity, controlling for the previously mentioned variables. The model was still statistically significant, χ2(2, N = 162) = 23.7, p < .02). Adding smile ratings led to a significant improvement in predicting mortality, χ2(2, N = 162) = 8.2, p < .017. Players with Duchenne smiles were half as likely to die in any year compared with nonsmilers, HR = 0.50, p = .006 (see Fig. 1), but Duchenne smilers did not differ significantly from partial smilers, who in turn did not differ significantly from nonsmilers. In this model, smile intensity accounted for 35% of the explained variability in survival (ratio of chi-squares: 8.2/23.7).
See, there’s no reason to be mad at Andruw.
Thanks to my colleague David Mitchell.