The previous chapter introduced the basics of utilizing reliability test, correlation coefficients, and multiple regression. This chapter presents the results of the data analysis. In the first section, the reliability analysis concerning the final version of the intention scale is presented. Secondly, the research offers the analysis of the relationship between the two language learning intentions and self-regulatory capacity via Pearson Product-Moment correlation coefficients. Finally, the analysis of hierarchical multiple regression to address to the research questions is displayed to explain the extent goal intentions/ implementation intentions can affect the demonstration of self-regulatory capacity for high school students in Taiwan.
The Reliability Test of the Intentional Constructs
It is important to note that alpha is a property of the scores on a test from a specific sample of testees. Investigators should not rely on published alpha estimates and should measure alpha each time the test is administered. The Cronbach’s alpha coefficients were thus computed to illustrate the reliability of intentional representations for the formal study. Table 8 summarizes the reliability of the indicator variables and the composite reliability of each surveyed construct. The results showed that all resulting coefficients exceeded the suggested threshold of .70 and even reached above .80, showing a significant internal consistency and convergent validity.
45
Table 8
The Internal Consistency of Intentional Subscales in the Formal Study
Latent Variables Subscales Item Number Cronbach’s
alpha Goal Intention
(.831)
Integrative Orientation 1, 4, 7, 9, 10, 11 .896 Instrumental Orientation 2, 3, 5, 6, 8 .800 Implementation
Intention (.915)
Content Orientation 2, 3, 4, 8 .826 Situational Orientation 9, 11, 12, 13 .852 Strategic Orientation 5, 6, 7, 10 .842 Interactive Orientation 1, 14, 15 .851
Although the reliability assessment was satisfactory for the formal study, the item-total statistics provided further analysis of the relationship among the statements.
Table 9 presents the item-total statistics of instrumental orientation, and a lower item-total correlation could be found in Item 8. In addition, while the overall internal consistency coefficient remained high (.80), the subscale would demonstrate even stronger consistency if Item 8 was deleted (.85). As a matter of fact, this particular item yielded a similar tendency in the pilot study. That is, the indicative power of this statement stayed relatively poor in the pilot and the formal study and thus is subject to modification (My goal in learning English is to achieve an “A” in the class).
46
Table 9
Item-total Statistics of Instrumental Orientation in the Formal Study Item proficiency can help me find a decent job.
.722 .722
I intend to learn English because I will need it for my future career.
.677 .736
G6
My goal in learning English is because it will assist me to apply for a better major.
.558 .770
G8
My goal in learning English is to achieve an “A”
in the class.
.361 .847
A similar index occurred in reporting the item-total statistics for strategic orientation. In the section of evaluating implementation intention, item 10 is considered to be less relevant with the subscale (corrected item-total correlation: .49).
Although the Cronbach’s alpha for strategic orientation revealed a strong uniformity among items (.84), a higher coefficient would be expected if item 10 was eliminated (.87).
47
Table 10
Item-total Statistics of Strategic Orientation in the Formal Study Item my concentration to make learning more effective.
.722 .780
If I encounter difficulties when I read English, then I have tactics to solve the problems and understand.
.713 .786
Table 11 shows the relationship between intentions and self-regulatory capacity (SRC). Integrative (r=.43) and instrumental orientation (r=.28) correlated significantly with SRC. In addition, content (r=.56), situational (r=.64), strategic (r=.82), and interactive orientations (r=.44) had a significant correlation with SRC as well. Since the independent variables were proved to be correlated with the dependent variables, the hierarchical regression analysis could be conducted to examine the predictive power of the predictor variables over the dependent variable.
Table 11 also demonstrates the discriminant validity among the variables.
Discriminant validity is to examine the extent to which a construct can be truly distinct from the other constructs. Significant discriminant validity means the
48
measured constructs are able to catch the unique characteristics that others do not. The correlation were computed to show whether the subscales can be discriminated from each other (correlations cannot be too close to 1.00, and <.85). The results show that the variables can be reasonably separated by the uniqueness each variable holds.
Table 11
Correlations Matrix among the Independent Variables and the Correlation Coefficients of Subscales with the Dependent Variable
1 2 3 4 5 6 7
The Effects of Intentions on Self-Regulatory Capacity
In the formal study, the overall interaction between intentions and self-regulatory capacity was analyzed with hierarchical multiple regression. In all analyses, the dependent variable was self-regulatory capacity. Predictor variables were added to the regression equation in two steps in that the effects of implementation intentions
49
should be based on the activation of goal intentions. The theoretical sets consist of subscales for goal intentions and implementation intentions. At step 1, the two control variables (integrative and instrumental orientations) were added, creating Model 1.
Next, all six variables were computed, creating Model 2. The results are presented in Table 12 and reported in an attempt to answer the two research questions.
Table 12
Results from Hierarchical Multiple Regression: Predicting Self-regulatory Capacities from Goal Intentions and Implementation Intentions
Model 1 Model 2 capacity for high school students in Taiwan?
50
In Table 12, the R² value for model 1 is .227, which means that the regression equation containing just two control variables accounted for about 23% of the variance in self-regulatory capacities, F(2, 386)= 56.648, p< .001. Standardized multiple regression coefficients (beta weights) for these two predictors indicate that both beta weights were significantly different from zero. Standardized multiple regression coefficients (beta weights) appear below the heading “β .” In Model 1, integrative and instrumental orientations indicate fairly significant explanatory power over self-regulatory capacity (β = .39, β = .21 respectively, p< .001). This result is in agreement with Gardner’s claim (1985) that integrative orientation is more critical in L2 learning.
2. To what extent can implementation intentions affect the demonstration of self-regulatory capacity for high school students in Taiwan?
Table 12 also demonstrates the percent of variability in the dependent variable that can be accounted for by all the predictors together. At Step 2, the four variables that constitute implementation intentions were added to the equation that contained the two control variables from Model 1. The R² value for the resulting model was R²=.712, F(6, 382)= 157.160, p< .001. Adding these four variables at Step 2 resulted in an increase in R² ofΔ R²=.485, F (4, 382)=160.579, p< .001. The change in R² is a way to evaluate how much predictive power was added to the model by the addition of another variable in step 2. In this case, the percent of variability accounted for self-regulatory capacity went up by almost 49%, which was considered much of an increase, in predicting self-regulatory capacities, beyond the variance already accounted for the two control variables in Model 1. This increase corresponds to an index of effect size of f²=1.68. According to criteria provided by Cohen (1988), this constitutes a large effect.
51
With respect to the four predictor variables that constitute implementation intentions, the results showed that standardized multiple regression coefficients for three of the four variables were significantly different from zero (p< .05), with the beta weight for situational orientation to be .17 (p< .001), that for strategic orientation to be the largest ( β=.66, p< .001), and that for interactive orientation to be .08 (p =.03) One of these four orientations in Step 2- content orientation- displayed a beta weight that was not significantly different from zero, β = -.023, p=.574. Another noteworthy finding in Model 2 is that the predictive strength reduced for goal intentions when the other four variables were added, with the beta weight of integrative orientation shrinking to insignificance (β = .50, p = .14) and the beta weight of instrumental orientation decreasing to .01(p=.001).
52