Studi Simulasi Model Nested Logit dan Paired Combinatorial Logit pada Respon Multinomial

Abstract

Multinomial logit model (MNL) is based on independent irrelevant alternatives assumption. If there was a correlation, MNL would result bias estimator. Correlation among of choices could be accommodated by Nested logit Model (NL) and Paired Combinatorial Logit Model (PCL). This research discuss how NL and PCL model can be used for estimating parameters (regression coefficient and correlation) and comparing to MNL in discrete choice model. Correlation structure is tested nesting correlation and correlation overlap. Paramaters in the models are estimated using maximum likelihood method. Simulated data generated using software R.3.0.1. Based on simulations data with nested correlation, NL model more appropriate than do MNL. But on data with overlapping correlation, the maximum likelihood estimator is not unique.