Using control charts in surveys

Mudryk, Burgess and Xiao (1996) propose to use statistical process control charts for computer aided telephone interview surveys at Statistics Canada. In this paper, we discuss two alternative charts to be used in the same application. By borrowing the concepts from hypothesis testing, we introduce criteria of coverage probability and detection power to compare the performance of these two charts. Both heuristic argument and empirical results show the equivalence of these two charts, Therefore, a choice is made based on relization convenience. In addition, we discuss how to adopt the ANOVA approach to testing the stability of an survey operation. The idea is tested with the weekly keying error rates observed during the twenty-one week data capture operation for the 1991 Census of Agriculture in Canada.