The seminonparametric (SNP) density estimation proposed by Gallant and
Nychka (1987) and Gallant and Tauchen (1989) has been applied in many
econometric analyses. In this paper, we show that the information mat
rix may become singular when an SNP model is overparameterized. This s
ingularity problem may occur even when a model selection criterion pen
alizes the size of a model, and thus cause problems in the sieves expa
nsion model selection process. We propose to use the likelihood ratio
test to safeguard against such overparameterization. (C) 1998 Elsevier
Science S.A.