Workpackage 3: Variance Estimation


The importance to develop an appropriate variance estimation methodology can be derived from many EU regulations for official statistics which foresee to publish several measures of accuracy which have to be elaborated with high quality. This can only be achieved when testing the methodology in a practical environment which is achieved within this work package via the interface to WP 6 the simulation study.

Measuring the accuracy of estimates from WP 2 generally leads to applying the appropriate variance estimation methodology. This allows for deriving standard errors, confidence intervals and other similar measures as well as of design effects which help measuring the impact of sampling designs on the outcome of the estimates. In general, two different methodologies can be applied, linearization and resampling methods. The efficiency of the variance estimators is mainly based the properties of the point estimators as well as of the possible skewness of the variable of interest. Further, outliers and rare populations may spoil the quality and appropriateness of the variance estimation methodology and may lead to other than usual conclusions which are of major practical impact.

The objective of the work package is to derive advanced recommendations for appropriate variance estimators of the Laeken indicators which take the different sampling schemes of the member states as well as the consideration of practical problems such as data peculiarities which may lead to an inappropriate use of indicator outcomes. This should allow for a more precise and valuable usage of indicators and hence for better use in policy making.