Some selection criteria for nested binary choice models: a comparative study (Q964651)

A generalized linear regression model is considered with binary responses. Three specification selection criteria are discussed: the Akaike information criterion (AIC), the Schwarz Bayesian information criterion (SBIC), and a new criterion, \(C_2\), is proposed by the authors. The \(C_2\) criterion is based on the least squares approach. The theoretical behaviour of the criteria is discussed. Their consistency is investigated. Efficiency and robustness of the criteria are studied via simulations.