| Summary: | 碩士 === 銘傳大學 === 財務金融學系碩士在職專班 === 99 === There is multilevel of information collected in this study, and when company’s dependent variable is influenced by industry level explanatory variable, we still take traditional regression analysis, the estimated standard errors will be too small and type I error over inflated; therefore, this study adopts to hierarchical linear modeling to analyze industry level, company level which will influence the company credit risk, and also compare both together with adding contextual variable aggregated by company level variable and high level explanatory variable in model and proceed with predicted ability and inspection of grade of fit from samples taken.
The result indicates there is 21.4% from total sum of variances of industry difference no matter company default or not, 78.6% from variation within-industry which means there is high connection between dependent variable and industry difference. Besides, no matter company default or not which causes cross-level effect could be through company level variable for contextual variable and high level explanatory variable; furthermore, putting contextual variable and high level explanatory variable in model, type I and type II error ratio are both reduced, and the accuracy of whole model is increased from 96.8% to 97%. The -2LL value of proper inspection becomes smaller and means the predicted test is more accurate.
|
|---|
