Bootstrap for Imputed Data in Panels: a Numerical Evaluation

by Lucia P. Barroso, Delhi T. P. Salinas .

Abstract: A common problem in sampling surveys is missing data. One way to handle this problem is using imputation techniques, which consists of predicting the missing observations to complete the data set which will be later analyzed. Imputation techniques can also be applied when the bootstrap method is used to estimate the standard error of some estimator. Although such techniques have been widely used, they have not received much attention for data sets with missing values. The main purpose of this paper is the application of non-parametric bootstrap techniques for data sets with missing values using imputation techniques to replace the missing values. We present the results of a simulation study about which deals with the estimation of the variance of the variance-components estimators in a mixed linear model. The samples we consider are panels, where $n$ individuals are observed in two occasions.

Key Words: BLUP, bootstrap, EM algorithm, imputation, mixed linear model, principal component analysis

Authors:
Lucia P. Barroso, lbarroso@ime.usp.br
Delhi T. P. Salinas, delhi@ime.usp.br

Editor: Roger Peck , rpeck@csub.edu

READING THE ARTICLE: You can read the article in portable document (.pdf) format (238634 bytes.)

NOTE: The content of this article is the intellectual property of the authors, who retains all rights to future publication.

This page has been accessed 2434 times since July 24, 2006.


Return to the InterStat Home Page.