A Variance Estimation R-package for Repeated Surveys – Useful for Estimates of Changes in Quarterly and Annual Averages

Abstract

This paper presents a newly developed R-package for calculation of variances of estimates based on data from several waves of a repeated survey with partly overlapping samples. Development of the package is a part of on-going work on quality improvements of the Labour Force Survey in Norway, which is quarterly and based on a rotating panel. The package can, for example, be used to calculate variances of net changes of annual averages of unemployment rates for persons aged 20-64. The methodology is based on linearly calibrated weights (as calculated by the packages ReGenesees and survey) and residuals from the corresponding regression modelling. These computations may be done separately for each wave. The functionality is generic and the user can specify any calibration model and any linear combination of (quarterly) estimates. Linearization is used to calculate variances of rates. The main method assumes that all relevant population totals can be computed from register data, but situations where totals are unknown for some of the calibration variables are also handled.

Keywords

• Calibration weighting
• Rotating panel survey
• Nonresponse adjustment
• Model-based
• Design-based
• Multivariate regression
• Linearization