Title: Estimating a semiparametric additive model for discrete choice data using backfitting algorithm
Authors: Patricia Gelin Doctolero - University of the Philippines Diliman (Philippines) [presenting]
Erniel Barrios - University of the Philippines (Philippines)
Joseph Ryan Lansangan - University of the Philippines (Philippines)
Abstract: Discrete choice models are estimated with the assumption of existence of a link function. To relax this assumption, a semiparametric additive model of the utility function is proposed. Quasi likelihood estimation is embedded into the backfitting algorithm and used in estimating the utility function. A simulation study is developed to evaluate the performance of the fitted model based on misclassification rate. The proposed model performed well in cases of linear and nonlinear nonparametric functions given all alternatives have balanced data allocation. Moreover, results showed that the model is robust to different magnitudes of misspecification error, and increases in sample size lead to slight increase in the predictive ability. Misclassification rates in validation data are also generally smaller than when using the usual generalized additive model.