Research project: Enhancing quality of business survey data
Activities under this thematic area are about investigating and developing statistical methodologies for producing high quality business statistics.
Activities under this thematic area are about investigating and developing statistical methodologies for producing high quality business statistics.
The focus of this WP is on developing statistical methodologies to deal with the unique statistical problems associated with the production and use of survey business data (e.g. distributions such as returns and turnovers, which are notably highly skewed and have many outliers), in the face of a rapidly increasing demand for disaggregate, small area and small group point and interval estimation methods in business survey estimates, which create unique problems in that, current sample sizes do not generally meet the necessary constraints for applying classical statistical estimation methods.
Activities under this thematic area are about investigating and developing statistical methodologies for producing high quality business statistics, which have unique problems since distributions of variables are typically highly skewed with many outliers. This implies that methods that might work well in certain areas, e.g. for social statistics, will not necessarily prove to be robust for the collection of business statistics. Henceforth, the focus of this WP is on variance estimation methods for business statistics and, specifically, problems associated to estimation and confidence intervals in business surveys; the impact and performance of single and multiple imputation methods on business statistics; variance and methods for the use of incomplete business data; new statistical methodologies for joining data from two sources (overlapping and disjoint units); and editing and imputation (focusing on outliers detection methods; as well as practical recommendations for their use).
In particular, this WP focuses on:
Deliverables include: an interim report and best practice recommendations on state-of-the-art estimation and variance estimation procedures and small area estimation methods for business statistics. To undertake this work we will generate synthetic micro data using sophisticated modelling techniques which might be considered for a dissemination strategy as a public-use file to researchers.