Nonresponse in sample surveys has been increasing over the years. This thesis covers that issue in two main parts. The first part is concerned with how to use observed data to make inference about regression coefficients in a linear regression model of cluster-level variables when some of the response variable data is missing. A naive approach estimates the regression coeffcients without considering nonresponse. We propose new methods for estimating coeffcients which incorporate information on nonresponse at the cluster level. We also extend Heckman estimators to our clustered model. The Workplace Employment Relations Survey (WERS) 2004 data and data from a prepared simulation study are used to compare the new methods with the naive approach. In the second part the generalized regression estimator (GREG) for two-stage sampling will be considered. We propose new optimum GREG estimators for stratified two-stage sampling and a simulation study is used in order to assess the performance of the new estimators.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:618776 |
| Date | January 2014 |
| Creators | Nangsue, Nuanpan |
| Contributors | Berger, Yves ; Skinner, Christopher ; Shlomo, Natalie |
| Publisher | University of Southampton |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | https://eprints.soton.ac.uk/366488/ |
Page generated in 0.0028 seconds