Evaluation of Logit and Probit Models in Mode-Choice Situation

Abstract

Logit and probit models are statistical tools that are well suited to analyze modal split situation. Logit is superior to probit from the analytical point-of-view, whereas the probit has more reliable theoretical basis. The main objective of this paper is to compare and evaluate the predictive ability of logit and probit models when applied in mode choice context. Two transit modes operating in different cities of the Saudi Kingdom were analyzed. The attributes of transport system, traveller, and trip were considered as they represent the major components of the utility function. Including transport system attributes was termed as a partial-specified model, while submitting traveller and trip attributes was denoted as a full-specified model. The evaluation stage was based on various criteria such as inconsistency and significance of model's coefficients, goodness-of-fit-measure, outlier analysis and market segment test. It is recommended to use full-specified model because of its ability to replicate the overall process of mode choice more accurately. Besides, the writer suggests not to calibrate the complicated models such as probit when analyzing the situation of a binary mode choice because logit model shows more accurate predictions.