A0946
Title: Confirmatory analysis to identify relative importance of regressors for the linear regression model
Authors: Tsung-Chi Cheng - National Chengchi University (Taiwan) [presenting]
Abstract: In various application analyses with regard to the linear regression models, the measurement of the individual relative importance of each explanatory variable in the model is of great practical significance and meaningfulness. Among the many approaches of relative importance, dominance analysis measures the relative importance of explanatory variables in an estimated model according to their contribution to the overall model fitting. Dominance analysis is quite intuitive, and its interpretation is very simple and straightforward. However, the so-called complete dominance is considered a purely qualitative comparison, while conditional dominance and general dominance are constructed in the context of quantitative concepts. Based on the framework of dominance analysis, the focus is the statistical hypothesis testing analysis for the confirmation of complete dominance, conditional dominance, and general dominance. The proposed methods are applied to construct the comprehensive measurement of subjective well-being (SWB) and identify those important factors affecting SWB.
