Comparison of the jackknife and bootstrap techniques for sample surveys with particular application to household surveys

Abstract

The complex estimators frequently used in modem surveys lead to particular difficulties in the estimation of sampling variance. This dissertation focuses on current methods of deriving such variance estimates. It considers two re-sampling methods, namely the Jackknife and the Bootstrap technique. This work combines the theoretical study of these two estimation techniques with a numerical application. A simulation study is carried out based on the Maltese Census 1995 and samples are extracted to mimic as close as possible the Labour Force Survey. The simulation study explores the variance estimates determined under the two techniques for a range of estimators. Consequently, this work is a reflection on the application of the above mentioned techniques as an attempt to extract the best technique for estimating the variance when applied to household surveys.