Weigted derivative estimation on quantal response models: Simulation and applications to choice of truck freight carrier (Q1424606)

Estimates of ratios of parameters in discrete choice (regression) single-index models using weighted average density derivative (WAD) estimators are discussed. The performance of two WAD estimators is studied in a small sample context and compared with the behaviour of a logit estimator by Monte-Carlo simulation. For spherical errors in a latent variable specification the WAD estimator, in terms of bias and mean square errors, demonstrates performance similar to the logit maximum likelihood estimator, while under heteroskedastic errors the WAD estimator performs better. Results of application of WAD estimator to a real data set on truck freight shipping firms from 2 different sectors (Food/Tobaco and Engineering) in Sweden are analysed.