Local Likelihood Estimation
Abstract
This paper extends the idea of local averaging to likelihood based models. One such application is to the class of generalized linear models. The author enlarges this class by replacing the covariate from chi beta with an unspecified smooth function. This function is estimated from the data by a technique called Local Likelihood Estimation - a type of local averaging. Multiple covariates are incorporated through a forward stepwise algorithm. The main application discussed however, is to the proportional hazards model of Cox (1972), for censored data, In a number of real data examples, the local likelihood technique proves to be effective in uncovering non-linear dependencies. Finally, the author gives some asymptotic results for local likelihood estimates and provides some methods for inference.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1984
- Accession Number
- ADA147317
Entities
People
- Robert Tibshirani
Organizations
- Stanford University