Estimating as semiparametric additive model for discrete choice data using backfitting algorithm

College

College of Science

Document Type

Dissertation

Publication Date

2018

Abstract

Discrete choice model is widely used in brand choice modelling and have assumptions of linearity in parameters. This assumption is relaxed through a semiparametric additive model of the utility function proposed in this study. Quasi likelihood estimation embedded in the backfitting algorithm was used to come up with the utilities. The alternative with the highest utility is chosen and the probabilities are computed. The performance of the model was evaluated through misclassification rate and results of the simulation studies with 3 categories shows that the postulated model’s performance is comparable with the nonparametric function specified either linearly or not. Also, the proposed model is robust to different magnitudes of misspecification error specifically 0.5,1 and 5. However, the model has the least preferred performance when subjected to unbalanced proportion of alternatives and with linear specification of the nonparametric function.

html

Disciplines

Statistics and Probability

Note

Creation date supplied

Keywords

Model theory

Upload File

wf_no

This document is currently not available here.

Share

COinS