A0633
Title: Estimation of threshold-boundary logistic regression models
Authors: ChihHao Chang - National Chengchi University (Taiwan) [presenting]
Abstract: The threshold boundary logistic regression (TBLR) model analyses binary data. The TBLR model combines logistic regression and threshold boundary functions using explanatory variables, allowing for the construction of the threshold boundary function by multiple explanatory variables to create linear or nonlinear classifiers. These classifiers split the binary data into two groups, and logistic regression models are fitted separately to each group. An ordered iterative algorithm named the TBLR-WSVM algorithm, which integrates weighted support vector machine (WSVM) and maximum likelihood estimation methods to estimate the TBLR model, is introduced. Simulation studies and empirical analyses are conducted to evaluate the performance of the TBLR-WSVM algorithm. Numerical analysis results demonstrate that the TBLR-WSVM algorithm exhibits robust estimation and prediction capabilities for linear and nonlinear threshold boundary logistic models, particularly under finite sample conditions.
