LR1_1 Effect of classification errors on domain level estimates in business statistics

Document date: 
Tuesday, 29 August, 2017

Estimation of bias and variance of the statistic of interest.

The method uses three basic steps. First classification errors are modelled, second the error sizes are estimated by collecting independent data and third the accuracy is estimated using a bootstrap approach. We consider the relatively simple case where classification errors are the only errors that affect the publication figures. This implies also that there are no sampling nor non-response errors, so we have values for all units in the population.

Document data sheet

Related content: