|Budget Amount *help
¥1,800,000 (Direct Cost : ¥1,800,000)
Fiscal Year 1995 : ¥100,000 (Direct Cost : ¥100,000)
Fiscal Year 1994 : ¥1,700,000 (Direct Cost : ¥1,700,000)
The last year of the project focused mainly on the reanalyzes of the whole data using the technique of multiple regression imputation of the missing values. Consequently, about one fifth of the cases that were formerly treated as missing was able to be saved from the archive and entered into the data set. However, the results obtained from this procedure did not show any noticeable difference compared with the previous results in terms of its best model selection, and the estimated parameters for the best model. Thus, the data set without such imputation was used for the final analyzes including exploring the prossibility of the improvement of the selected best model in order to avoid potential problems causedby data imputation, such as linear dependency, estimates being fixed to lower and/or upper boundaries.
One important finding of the current study was that the Equal Influence Model, which assumed mutual correlational links among all the learner background factors including Monitor turned out to be significantly better than other models, which presumed Monitor was not an independent but a dependent factor generated by formal, explicit rule oriented instructions. This result supported the models assuming the interface between Krashen's Learning and Acquisition. In particular, Monitor and other factors, especially, Informal Exposure to English had significant correlations. This result suggests the possibility of a new and theoretically interesting model, which might be called "Reversed Interface Model". In this model, Monitor could affect the Informal Exposure to English as well as the other way round