C. Structural Model Assessment
2. Hypothesis Test
The current research employed a structural equation modeling (SEM) technique for testing hypothesis path. According to Jöreskog and Sörbom (1989), with increasing complexity and specificity of social behaviors and appearance of flexible, user-friendly computer software in econometrics, SEM has come to the fore as a standard approach to analyzing various social phenomena. Hoyle (1995) and Kline (1998) suggested that SEM is a comprehensive, flexible, and analytical statistical tool for testing hypotheses about relationships among observed data and latent variables. The advantages of SEM are reducing biases with measurement errors, better estimating path coefficients, ability to estimate both direct and indirect effects, and ultimately resulting in a model fit.
Additionally, this strict test has an absolute advantage over the model with background of previous theories, just as the present study.
The SEM test yields some important information such as Standard Error (S.E.), Critical Ratio (
t
-value), and standard path coefficient (β
).t
-value of individual parameters is computed by S.E. in order to test the null hypothesis that path value is equal to zero (H0:β
=0). At
-value greater than ± 1.96 threshold refers to the probability level (p
) of 0.05, interpreted as significantly different from zero. And, if it is bigger than ± 2.58, the path has the level 0.01 of probability (p
). Furthermore, standard path coefficient (β
) is the most crucial criteria used to decide whether or not causal relationships between variables are valid (Wixom & Watson, 2001), which will in turn explain the results for research hypothesis. In other words, abnormality of the path coefficients is completely determined by the standard coefficient value. In that case of the standard coefficient value above '1', a new path analysis should be carried out with a modified model. The results of path analysis in the structural model are presented in Table 13.RH Path S.C.(β) N.C. Std.
Error
t-value
(Critical Ratio) Results
RH 1 MOT ⇨ ACH 0.275 0.279 0.093 3.033** Accepted
RH 2 ANX ⇨ ACH 0.137 0.141 0.085 1.659 Rejected
RH 3 ATT ⇨ ACH 0.252 0.256 0.087 2.943** Accepted
RH 4 PEOU ⇨ ACH 0.122 0.126 0.084 1.500 Rejected
RH 5 PU ⇨ ACH 0.219 0.225 0.081 2.778** Accepted
RH 6 PI ⇨ ACH 0.212 0.217 0.091 2.385* Accepted
RH 7 VAR ⇨ ACH 0.109 0.113 0.085 1.329 Rejected
[Table 13] Results of Hypothesis Test
Note. MOT: Motivation; ANX: Anxiety; ATT: Attitude; PEOU: Perceived Ease of Use, PU;
Perceived Usefulness; PI: Perceived Interactivity; VAR: Variety; ACH: Academic Achievement; S.C.(β): Standard-path Coefficient; N.C.: Nonstandard Coefficient. p<0.05:*, p<0.01:**
As exhibited in Table 13, the
t
-values of three paths from motivation (3.033), attention (2.943), and perceived usefulness (2.778) toward achievement are greater than ± 2.58, standing at the high probability level (p
) of 0.01. And thet
-value of perceived interactivity toward achievement is over ± 1.96 at the probability (p
) of 0.05 level, meaning statistically significant estimate. However, thet
-values of anxiety, perceived ease of use, and variety do not show probable effects on the paths toward achievement. Now, standard path coefficient (β
) is examined for testifying the measurement hypotheses as follows:[RH 1] MOT ⇨ ACH : With a path coefficient of 0.275 pointing at the p
< 0.01 level of significance, it is found that EFL learners' motivation in the context of blended learning has a significant effect on their achievement. Obviously, it demonstrates the hypothesis 1 that there is a robust relationship between two of them.
[RH 2] ANX ⇨ ACH : A correlation coefficient of 0.137 refers to n.s;
i.e., there is 'no significance' between EFL learners' anxiety and their academic achievement in the context of blended learning. Therefore,
the hypothesis 2 is not accepted.
[RH 3] ATT ⇨ ACH : EFL learners' attention toward achievement is found as a path coefficient of 0.252 value by the 0.01 level of significance, which acknowledges the factor of attention toward blended learning use is a determinant and in turn backs up the hypothesis 3.
[RH 4] PEOU ⇨ ACH : The perceived ease of use of the blended learning system, with 0.122 correlation, does not have a direct effect on learners' English academic achievement, which means the hypothesis 4 is rejected. Apparently, the perceived ease of use hardly affects achievement of L2 in any way.
[RH 5] PU ⇨ ACH : It is identified that the perceived usefulness of the blended learning system was highly of relevance in learners' English academic achievement. A correlation coefficient of 0.219 corresponds to the significance level of p < 0.01, making the hypothesis 5 supported.
[RH 6] PI ⇨ ACH : The perceived interactivity in blended learning shows its statistical relationships of the p < 0.05 level of significance with English learners' academic achievement, indicating a correlation coefficient of 0.212 and making the hypothesis 6 acceptable.
[RH 7] VAR ⇨ ACH : A significant effect of contextual variety of blended learning on achievement is not observed to inform an answer to the hypothesis 7. A path coefficient of 0.109 cannot be proof of the path, and to no effect.
These tested hypotheses are schematized in the following path-diagram.
Figure 10 visualizes paths of the variables in the structural model along with the standardized path coefficients.
[Figure 10] Path-Diagram of Tested Research Hypothesis
In the affective domain, Hypotheses 1 (motivation toward achievement) and 3 (attention toward achievement) were verified; out of the contextual factors, Hypotheses 5 (perceived usefulness toward achievement) and 6 (perceived interactivity toward achievement) were identified. And the other paths (anxiety toward achievement, perceived ease of use toward achievement, and variety toward achievement) were not found out to be significant. Considering the standard path coefficient of .275, influence of motivation on achievement has the biggest explanation, and impact of attention on achievement is also fairly large. But, variety has the least impact on achievement with the path coefficient of .109. To wrap it up, the higher exogenous variables of motivation and attitude, perceived ease of use, and perceived interactivity, the higher an endogenous variable of academic achievement in blended learning.