316 Chapter 7 Table 5 Results from a binary logistic regression predicting game choice from emotion and stress mindsets Predictor B SE B Wald’s χ2 df p eB (odds ratio) Constant 0.52 0.99 0.28 1 .60 1.68 Experimental conditiona 1.80 0.41 19.67 1 < .001 6.06 Emotion mindset pre-test -0.16 0.26 0.37 1 .54 0.85 Stress mindset pre-test -0.30 0.34 0.77 1 .38 0.74 Test χ2 df p Omnibus test of model coefficients 22.27 3 < .001 Goodness-of-fit test Hosmer & Lemeshow 3.28 8 .92 Notes. Dependent variable ‘game choice’ was coded as 0 = entertainment, 1 = mental health. -2 Log likelihood = 151.69. Cox and Snell R2 = .16. Nagelkerke R2 = .21. When gender was included as a control variable, similar results were found and gender was not a significant predictor. a coded as 0 = detailed trailer with mental health messaging, 1 = abstract trailer with mental health messaging. Changes in Emotion and Stress Mindsets based on Game Choice To examine whether emotion and stress mindsets changed when participants chose the mental health trailer (and not the entertainment trailer), two Repeated Measures ANOVA’s were performed. One analysis included pre-test and post-test emotion mindset as a within-subjects factor. The other analysis included pre-test and post-test stress mindset as a within-subjects factor. In addition, both analyses included game choice as a between-subjects factor and experimental condition as a covariate. Statistical assumptions were tested and met1. Overall, emotion mindset did not change from pre- to post-test, F(1, 126) = 1.68, p = .20, η2 p =.01. In addition, there was no significant interaction effect between game choice and the change in emotion mindset, F(1, 126) = .73, p = 1 Box’s M test of equality of covariance matrices was significant in the Repeated Measures ANOVA for emotion mindset. Tabachnick and Fidell (2014) state that this test is highly sensitive and that only if the sample sizes are unequal and Box’s M test is significant at p < .001, robustness is not guaranteed. Our sample sizes were unequal, but the p-value of the test was .03 suggesting the test is still robust. In addition, it has been suggested to use Pillai’s criterion instead of Wilks’ lambda to evaluate multivariate significance when covariance matrices are unequal (Olson, 1979). However, Pillai’s criterion provided the same results as Wilks’ lambda in our analysis. Therefore, we ignored the significance of the Box’s M test.
RkJQdWJsaXNoZXIy MTk4NDMw