Construction of mathematical models in language learning strategy and English proficiency of pre-university students
This study aims to construct the best mathematical model that describes the relationship between six Language Learning Strategies and English proficiency of the pre-university students of Universiti Malaysia Sabah. Two hundred and thirty pre-university students of Universiti Malaysia Sabah took...
محفوظ في:
| المؤلف الرئيسي: | |
|---|---|
| التنسيق: | أطروحة |
| اللغة: | English |
| منشور في: |
2016
|
| الموضوعات: | |
| الوصول للمادة أونلاين: | https://eprints.ums.edu.my/id/eprint/12754/1/Construction%20of%20mathematical%20models.pdf |
| الوسوم: |
إضافة وسم
لا توجد وسوم, كن أول من يضع وسما على هذه التسجيلة!
|
| الملخص: | This study aims to construct the best mathematical model that describes the
relationship between six Language Learning Strategies and English
proficiency of the pre-university students of Universiti Malaysia Sabah. Two
hundred and thirty pre-university students of Universiti Malaysia Sabah took
part in this study by answering a background questionnaire and the Strategy
Inventory for Language Learning self-report questionnaire. Also, these
students sat for the Malaysian University English Test (MUET) and their
results are taken as the student's proficiency in English as a second
language. Three linear regression models was used as initial models, two of
which went through stepwise variable selection method. However, in this
study, we proposed two families of nonlinear models, which are the
nonlinear Gompertz model, and a modified Gompertz model. Based on the
five proposed models, the dependent variable, y, is the normalized English
proficiency of the students and the independent variables, Xj , are the six
different language learning strategies where j = 1,2,3, ... ,6. These six
strategies are memory, cognitive, comprehensive, metacognitive, social and
affective respectively. Goodness-of-fit tests and information criterion were
used to assess these models, which are the root mean square error (RMSE),
mean absolute error (MAE), residual standard error (RSE), the corrected
Akaike information criterion (AlCe) and Bayesian information criterion (BIC).
The results were then used to compare these models. The linear model
showed good readings with its RMSE, MAE and RSE approaching zero
indicating that the model is strong. However, the Gompertz model and the
modified Gompertz model showed even better values of RMSE, MAE, RSE
and AlCe. Especially for the modified Gompertz model, the values for the
errors were even closer to zero. The goodness-of-fit tests and AlCc show
that the modified Gompertz model's results were better than the nonlinear
Gompertz model, and the linear regression model. Despite that, only the BIC
calculated for the linear model M35 has the weightiest amount of
information. In conclusion, four out of five metrics calculated suggested that
the modified Gompertz model is the best amongst investigated models. |
|---|