Inhibition and PRA. I predicted that Inhibition-Negative correlates with PRA,
but that Inhibition-Neutral and Inhibition-Positive do not. I conducted three hierarchical multiple regression analyses to test this hypothesis and a Bonferroni adjusted alpha level of .0167 was applied. Each PRA was entered as the dependent variable; emotion
reactivity (ER) of the same emotion as PRA was entered as the predictor variable in the first step and inhibition of valence-specific information was entered as the predictor in the second step. Inhibition-Neutral, Inhibition-Positive, and Inhibition-Negative were examined in three separate regression models.
Contrary to my prediction, Inhibition-Negative was not correlated with three types of PRA, all ps >.05 (see Table 5.3). As predicted, no association between Inhibition-Neutral and PRA was found, all ps >.05 (see Table 5.1); no association between Inhibition-Positive and PRA was found, all ps > .05 (see Table 5.2).
WM and PRA. I predicted that WM-Positive positively correlates with PRA, but
WM-Neutral and WM-Negative do not. I conducted three hierarchical multiple
regression analyses to test this hypothesis and a Bonferroni adjusted alpha level of .0167 was applied. Each PRA was entered as the dependent variable; ER for the same emotion of PRA was entered as the predictor in the first step and the valence-specific WM was
entered as the predictor in the second step. WM-Neutral, WM-Positive, and WM- Negative were examined in three separate regression models.
Contrary to my prediction, WM-Positive was not associated with three types of PRA, all ps >.05 (see Table 6.2). As predicted, no association between WM-Neutral and PRA was found, all ps >.05 (see Table 6.1); no association between WM-Negative and PRA was found either, all ps > .05 (see Table 6.3).
Secondary analyses 1– nonlinear relationships. The main analyses indicated no
linear relationship between valence-specific EF and PRA; however, it is possible that the relationships are not linear. Thus, I conducted secondary analyses to explore the
possibility of curvilinear relationships between valence-specific EF and PRA.
Specifically, I examined the quadratic relationships between valence-specific EF and PRA because it is possible that as valence-specific EF increases to a level that supports the highest level of PRA, the increase in PRA may level off.
Inhibition and PRA. To examine the quadratic relationships between PRA and inhibition, I conducted three hierarchical multiple regression analyses and a Bonferroni adjusted alpha level of .0167 was applied. Each PRA was entered as the dependent variable. ER of the same emotion as PRA was entered as the predictor in the first step; inhibition of valence-specific information was entered as the predictor in the second step; inhibition squared was entered as the predictor in the third step. Inhibition-Neutral, Inhibition-Positive, and Inhibition-Negative were examined in three separate regression models. Results showed no association between inhibition of valence-specific
WM and PRA. To examine the quadratic relationships between PRA and WM, I
conducted three hierarchical multiple regression analyses and a Bonferroni adjusted alpha level of .0167 was applied. Each PRA was entered as the dependent variable. ER of the same emotion as PRA was entered as the predictor variable in the first step; WM for valence-specific information was entered as the predictor in the second step; WM squared was entered as the predictor in the third step. WM-Neutral, WM-Positive, and WM- Negative were examined in three separate regression models. Results showed no association between WM for valence-specific information and PRA all ps > .05 (see Table 8.1-8.3).
Secondary analyses 2– extreme groups approach. The analyses so far have
shown that no linear or curvilinear relationship exists between valence-specific EF and PRA. The next analytic strategy was to sample extreme values of valence-specific EF. This methodology is called extreme groups approach (EGA). This is not a redundant step after curvilinear regression analyses because EGA captures extreme scores whereas curvilinear regression analysis concerns the shape of the relationships of two variables. EGA has been widely used in studies that investigate EF in college students to ensure that the lower end of the EF distribution was not under represented (e.g., Schmeichel et al., 2008) and to enhance statistical power (Kelly, 1939). In the current study, ninety-three percent of the participants had an education level equal to or higher than the associate degree or partial college. It is possible that the lower extreme values of the EF were under represented in the sample. In other words, sampling from the extremes ensures that the key variables in my hypothesis were adequately sampled, as suggested by Schmeichel et al. (2008). However, I have to note that the use of EGA could inflate the magnitude of a
relationship (Preacher, Rucker, MacCullum, & Nicewander, 2005) and the results need to be interpreted with caution.
Inhibition and PRA. To examine the relationships between PRA and inhibition of
valence-specific information by EGA, I conducted three independent groups t tests and a Bonferroni adjusted alpha level of .0167 was applied. Three types of standardized PRA residuals (controlling for the variance from the corresponding emotion reactivity) were entered as test variables and two extreme groups (high vs. low; individuals who scored higher than 1 SD above the mean on inhibition scores belong to the high extreme group, and individuals who scored lower than 1 SD below the mean on inhibition scores belong to the low extreme group) were entered as the grouping variable. Inhibition-Neutral, Inhibition-Positive, and Inhibition-Negative were examined in three separate sets of t tests.
Results showed that individuals with the extreme inhibition scores do not differ on PRA. For Inhibition-Negative, group differences on PRA-Happiness, PRA-Love, and PRA-Optimism were not significant, all ps > .05; similarly for Inhibition-Positive and Inhibition-Neutral, group differences on PRA were not significant, all ps > .05. Overall, the results in EGA revealed the same non-significant relationship as in regression
analyses such that inhibition of valence-specific information did not correlate with PRA.
WM and PRA. To examine the relationships between PRA and WM for valence-
specific information by EGA, I conducted three independent groups t tests and a Bonferroni adjusted alpha level of .0167 was applied. Three types of standardized PRA residuals (controlling for the variance from the corresponding ER) were entered as test variables and two extreme groups (high vs. low; individuals who scored higher than 1 SD
above the mean on WM scores belong to the high extreme group, and individuals who scored lower than 1 SD below the mean on WM scores belong to the low extreme group) were entered as the grouping variable. WM-Neutral, WM-Positive, and WM-Negative were examined in three separate sets of t tests.
Results showed that individuals with the extreme WM scores do not differ on PRA. For WM-Positive, group differences on PRA-Happiness, PRA-Love, and PRA- Optimism were not significant, all ps > .05; similarly for WM-Neutral and WM-
Negative, group differences on PRA were not significant, all ps > .05. Overall, the results in EGA revealed the same non-significant relationship as in regression analyses, such that WM for valence-specific information did not correlate with PRA.
Summary. I examined Question one (the relationships between valence-specific
EF and PRA) by employing three statistical methods (linear regression analysis, curvilinear regression analysis, and extreme groups analysis). Three methods have yielded similar results. Contrary to my prediction, PRA was not associated with Inhibition-Negative or WM-Positive, either in a linear or quadratic way.