The Taylor linearization and Jackknife techniques : theory and application to EU-SILC poverty indicators

Abstract

The continuous increase in interest to improve the quality of data from sample surveys is being emphasized in all statistical agencies around Europe. However, assessing the accuracy of estimates derived from a sample survey is generally not easy, due to the complexity in the nature of the sample survey design or in the nature of the statistics under study. This dissertation focuses on two variance estimation techniques which may help in improving the accuracy of the estimates. These are the Taylor Linearization and the Jackknife techniques. The performance of these two techniques and also the performance of two variations of the jackknife, namely the delete-1 and the delete-d jackknife techniques, are studied on the basis of the Survey on Income and Living Conditions (SILC) for the year 2005, for two statistics, namely the At-Risk-of-Poverty Rate and the Quintile Share Ratio. This is done with the attempt to find out which technique gives the best contribution to improving the data quality of household sample surveys, such as the SILC