G-TELPからTOEICスコアを予測する回帰モデルの検証 : 2年間のデータから示唆されること  [in Japanese] Regression Model Predicting the Scores of TOEIC from the G-TELP : Suggestions and Implications Obtained from the Data for the Two Years  [in Japanese]

The purpose of this study is to investigate the possibility of predicting the scores of TOEIC tests from the scores of G-TELP Level 3 (General Tests of English Language Proficiency), by using data collected from April 2013 to July 2013. To obtain the predicting model, the data collected from all first-year students of Nagasaki University in 2013 (N=1,388) were submitted for simple regression analyses. Then, to predict the TOEIC scores more precisely, the scores of each section of G-TELP test (Grammar, Listening, Reading & Vocabulary sections) were analyzed using the multiple regression analyses. These models obtained from the 2013 data were compared with the ones obtained from the 2012 data (Ogasawara & Maruyama, 2014). Although R^2 is slightly lower than in 2012 (R^2=.56 vs. R^2=.52), the model obtained from the 2013 data also indicates that the overall fit of G-TELP Level 3 scores to TOEIC scores is sufficient but the fit among the higher scores is not adequate due to the fact that G-TELP Level 3 does not differentiate well enough TOEIC scores of 600 and higher. To ascertain the higher R^2, a different kind of technique was used. Forty students from the nine faculties were randomly sampled and their scores of the TOEIC test and G-TELP test were analyzed by the multiple regression analyses as well as the simple regression analyses. The model obtained from this random sampling technique also indicates the overall fit of G-TELP Level 3 scores to TOEIC scores is sufficient, except for the TOEIC scores of 600 or higher and of 400 or lower. Moreover, R^2 is higher than one of the first model (R^2=.59 vs. R^2=.52).