Estimation of a partially linear additive model for data from an outcome-dependent sampling design with a continuous outcome.
Biostatistics
; 17(4): 663-76, 2016 10.
Article
em En
| MEDLINE
| ID: mdl-27006375
ABSTRACT
Outcome-dependent sampling (ODS) designs have been well recognized as a cost-effective way to enhance study efficiency in both statistical literature and biomedical and epidemiologic studies. A partially linear additive model (PLAM) is widely applied in real problems because it allows for a flexible specification of the dependence of the response on some covariates in a linear fashion and other covariates in a nonlinear non-parametric fashion. Motivated by an epidemiological study investigating the effect of prenatal polychlorinated biphenyls exposure on children's intelligence quotient (IQ) at age 7 years, we propose a PLAM in this article to investigate a more flexible non-parametric inference on the relationships among the response and covariates under the ODS scheme. We propose the estimation method and establish the asymptotic properties of the proposed estimator. Simulation studies are conducted to show the improved efficiency of the proposed ODS estimator for PLAM compared with that from a traditional simple random sampling design with the same sample size. The data of the above-mentioned study is analyzed to illustrate the proposed method.
Palavras-chave
Texto completo:
1
Base de dados:
MEDLINE
Assunto principal:
Projetos de Pesquisa
/
Interpretação Estatística de Dados
/
Modelos Teóricos
Tipo de estudo:
Prognostic_studies
Limite:
Child
/
Female
/
Humans
/
Pregnancy
Idioma:
En
Revista:
Biostatistics
Ano de publicação:
2016
Tipo de documento:
Article
País de afiliação:
China