Data Presentation Standards for Proportion from Complex Sample Survey

The precision evaluation and data presentation standards for sample estimates from complex sample survey is often based on the relative standard error. This is because most surveys include a wide range of characteristics, and it is preferable to use a standardized measure of sampling error that is independent of the unit of estimation. However, the relative standard error is not an appropriate criterion for evaluating the precision and data presentation standards due to its limitation for proportion in particular by definition. The relative standard error can have excessively large for an extremely large proportion. Recently, the U.S. National Center for Health Statistics has proposed an alternative data presentation standard for proportion based upon confidence intervals that reflect the design components of a complex sample design. Under an adequate effective sample size, confidence interval width and/or its relative width are used to determine whether sample proportion estimates can be presented or not. This paper discusses about the data presentation standards for proportion from complex sample survey recently proposed by U.S. National Center for Health Statistics (NCHS) and examines the prevalence estimates of mental disorders from the 2021 Mental Health Survey to compare whether they can be determined presentable to the public according to the NCHS and the Statistics Korea’s data presentation standards. The study results shows that confidence interval-based criteria should also be considered for improving Statistical Korea’s data presentation standards for proportion. suggest the need for the new NCHS’ approach taken into consideration