Testing: Likelihood Ratio Test

This lecture introduces the likelihood ratio test in the context of choice models, comparing unrestricted and restricted models with equal probability and alternative specific constants. It revisits benchmarking, discussing log likelihood statistics and classical output of estimation software. The instructor explains tests for generic attributes, taste variations, and nonlinear specifications, including power series and linear models. The lecture concludes with a caution against overfitting and the use of other nonlinear specifications like piecewise linear and Box-Cox transformations.