A carregar...
Publicação
Likelihood-based Inference for Multivariate Regression Models using Synthetic Data
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 1003.55 KB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
Likelihood-based exact inference procedures are derived for the multivariate
regression model, for singly and multiply imputed synthetic data generated
via Posterior Predictive Sampling (PPS), via a newly proposed sampling method,
which will be called Fixed-Posterior Predictive Sampling (FPPS), and via Plug-in
sampling. By contemplating the single imputation case, the new developed procedures
fill the gap in the existing literature where inferential methods are only
available for multiple imputation and, by being based in exact distributions, it
may even be applied to cases where the sample size is small. Simulation studies
compare the results obtained from all the proposed exact inferential procedures
and also compare these with the results obtained from the adaptation of Reiter’s
combination rule to multiply imputed synthetic datasets. An application using
U.S. 2000 Current Population Survey data is discussed and measures of privacy
are presented and compared among all methods.
Descrição
Palavras-chave
Finite sample inference Maximum likelihood estimation Pivotal quantity Partially synthetic data Statistical Disclosure Control Unbiased estimators