Merging information for semiparametric density estimation
Date
2004ISSN
1369-7412Source
Journal of the Royal Statistical Society.Series B: Statistical MethodologyVolume
66Issue
4Pages
941-958Google Scholar check
Keyword(s):
Metadata
Show full item recordAbstract
The density ratio model specifies that the likelihood ratio of m - 1 probability density functions with respect to the mth is of known parametric form without reference to any parametric model. We study the semiparametric inference problem that is related to the density ratio model by appealing to the methodology of empirical likelihood. The combined data from all the samples leads to more efficient kernel density estimators for the unknown distributions. We adopt variants of well-established techniques to choose the smoothing parameter for the density estimators proposed.