The purpose of Study 4 is to test two hypotheses that grit mediates the relationship between self-perception and academic performance. One hypothesis referred that grit was a significant mediator between self-perception bias and academic performance. The other hypothesis is that grit was a significant mediator between self-perception intentionally and academic performance.
6.1 Method
6.1.1 Participants
The participants in study 4 are the same as study 1.
6.1.2 Procedure
Before administering the study in Taiwan, we used a translation and back-translation process for Short Grit Scale (Grit–S). A researcher translated the measures into Chinese. Then, a second person back-translated the measures into English. Differences between the original English and the back-translation were discussed, and mutual agreements were made as to the most appropriate translation.
The first step is identical to that in Study 1. First, all participants took 15 min to complete 20 math questions and were asked to report the estimated number of questions that they had answered correctly. Second, participants were asked to complete some psychological questionnaire which included the Short Grit Scale (Grit–S). Then, participants were asked to estimate their performance again, this time with a monetary incentive.
65
6.1.3 Measures
6.1.3.1 Short Grit Scale (Grit–S, (Duckworth & Quinn, 2009)
The Short Grit Scale (Grit–S) measures trait-level perseverance and passion. This is a 6-item version constructed in response to improved psychometric properties of the original measure. The revised scale had adequate internal consistency (alpha = 0.60~0.79), and the goodness-of-fit indexes of confirmatory factor model is suitable for the two-factor model of grit. Among adolescents, the Grit–S longitudinally predicted GPA.
6.1.3.2 Measuring self-perception score
The same as Study 3, two kinds of residual discrepancies were used to test the hypothesis respectively. These were adjusted residual discrepancy and residual discrepancy-change. The different ways of calculating self-perception via these two discrepancies are elaborated as below.
First, when self-perception bias is measured by adjusted residual discrepancy, we need to take the absolute value of self-perception bias. The reason is that there should be a quadratic effect between self-perception and academic performance from the results of Study 1. Thus, in order to distinguish accurate assessors unintentionally from people either self-enhancing or self-effacing unintentionally, the data was modified by taking the absolute value of adjusted residual discrepancy. Those with a score of zero were accurate self-assessors unintentionally.
Those with positive residuals were self-enhancers or self-effacers unintentionally. Then, the absolute value could be used to predict the mediator.
Second, when self-perception is measured by residual discrepancy-change, we do not need to take the absolute value of self-perception. The reason is that there should be a linear effect between self-perception intentionally and academic performance from the results of Study 1.
Those with positive residuals were self-effacers intentionally, those with negative residuals were
66
self-enhancers intentionally, and those with a score of zero were accurate self-assessors intentionally.
6.1.3.3 Dependent Variable—academic performance
We recorded data for a variety of academic performance variables from school records. These variables included teacher assessments of everyday performance (50%) and three exams (50%) of all academic subjects in the Spring semester of 2011.
6.2 Result
According to Baron and Kenney’s recommendation(1986) and supplemented by the Sobel test (Preacher & Hayes, 2004), three criteria must be met for a variable to be considered a mediator: 1) the independent (self-perception)variable must predict the mediator (grit). 2) the independent variable (self-perception) must predict the dependent variable (academic performance), and 3) the mediator (grit) must predict the dependent variable(academic performance) in the multiple regression equation, with a reduced effect of the independent variable(self-perception) on the dependent variable after controlling for the mediator.
In study 4, two kinds of residual discrepancies were used to represent self-perception and test the hypothesis respectively.
6.2.1. Adjusted Residual Discrepancy: Grit was a significant mediator between self-perception bias and academic performance
When self-perception bias measured by adjusted residual discrepancy was used as the independent variable, the mediation effect was inspected following the regression analysis as outlined in Baron and Kenny (1986). First, self-perception bias was significantly related to academic performance (β = -.17 , p< .05). Second, self-perception bias was significantly related to grit ( β = -.17, p< .05). Third, grit was significantly associated with academic performance in
67
the multiple regression equation (β = .26 , p< .001). with a reduced effect of self-perception bias variable. The effect of the self-perception bias on academic performance became no longer significant and the effect of the self-perception bias was reduced after controlling for grit. Finally, the results of the Sobel test indicated that grit was a significant mediator for academic performance (Z=-2.10, p < .05). Table 6.2a summarizes results from the analysis, and Figure 6.2a shows these results as diagrams representing the mediated model.
Table 6.2a
Regression Analysis to Test Mediation Effect of Grit on Academic Performance (self-perception measured by adjusted residual discrepancy)
R2 β t p
Step 1: Regression of self-perception bias on academic performance
.03 -.17 -2.45 .015
Step 2: Regression of self-perception bias on Grit .03 -.17 -2.46 .015 Step 3: Multiple regression on academic performance
Self-perception bias
Sobel Test of Mediation Analysis Z=-2.10, p =.036*
Note. * p < .05. *** p < .001.
68
Figure 6.2a Path diagram of the mediation effect of grit on academic performance.
(self-perception measured by adjusted residual discrepancy) Values on paths are standardized β. * p < .05. *** p < .001.
6.2.2. Residual Discrepancy-Change: Grit was not a significant mediator between self-perception intentionally and academic performance
When self-perception intentionally measured by residual discrepancy-change was used as the independent variable, the mediation effect was inspected following the regression analysis as outlined in Baron and Kenny (1986). First, self-perception intentionally was significantly related to academic performance (β = .20, p< .01). Second, self-perception intentionally was not significantly related to grit ( β = .00, ns). That is, the effect of the mediation was not significant because the second equation didn’t fit Baron and Kenney (1986) ’s criteria. Table 6.2b summarizes results from the analysis, and Figure 6.2b shows the results as diagrams testing the mediation effect of Grit on Academic Performance (self-perception measured by residual discrepancy-change).
69
Table 6.2b
Regression Analysis to Test Mediation Effect of Grit on Academic Performance (self-perception measured by residual discrepancy-change)
R2 β t p
Step 1: Regression of self-perception intentionally on academic performance
.04 .20 2.95 .004
Step 2: Regression of self-perception intentionally on Grit
.00 .00 0.06 .95
Step 3: Multiple regression on academic performance Self-perception intentionally
Sobel Test of Mediation Analysis Z=0.08, p =.013
Note. * p < .05. ** p < .01. *** p < .001.
Figure 6.2b Path diagram of the mediation effect of grit on academic performance.
(self-perception measured by residual discrepancy-change) Values on paths are standardized β. * p < .05. ** p < .01. *** p < .001.
70