Use of indicators from data editing for monitoring the quality of the survey process: the Italian information system for survey documentation (SIDI)

Abstract

Information about the presence of non-sampling errors affecting survey data is of importance both for final users and for survey managers. Besides the direct methods for evaluating the effects of non-sampling errors on survey estimates, the analysis of data obtained from each phase of the production process represents an easy and inexpensive way to obtain information about quality. In particular in this work we would like to focus the attention on the editing phase and its relationships with the other phases of the production process. In fact, the amount of errors detected during the editing and imputation phase can provide information about the non-sampling errors affecting the previous phases of the production process. On the other hand, the editing and imputation procedures should be carefully defined and tested before their application on actual data in order to avoid overediting or other unpleasant “side-effects”. Given the relationship between non-sampling errors, it is important to jointly analyse quality indicators coming from the different phases of the survey production process. To this purpose, ISTAT has developed an information system devoted to embody survey metadata and quality indicators for each of the process phases. This information system enables both to monitor quality indicators over time for a single survey and to compare quality indicators among homogeneous surveys. A set of these quality indicators related to editing and imputation activity are described in this work and a strategy for their interpretation is also presented. Finally an application of these indicators to the Italian Labour Force survey is shown.