Abstract: | The results of a simple counting experiment are presented and analyzed. A model is set up that includes bias and intra-enumerator correlation; these components are estimated and their effect projected for a more general sampling situation. This provides a realistic data set to give greater insight into the nature and prevalence of measurement error. In this example the measurement bias is about four percent, with a .13 intra-enumerator correlation. This experiment is suitable for a classroom exercise for introductory statistical methods or for more advanced statistics students. |