The Control Charts as a Quality Tool for Monitoring C.A.T.I. Surveys

The objective of this paper is to show how to use control charts for monitoring the interviewers’ work in CATI periodic surveys. For data collection of the second edition of Survey on Citizens’ Safety (an ISTAT periodic CATI survey) we are using unvariate “Shewhart charts for attribute” for monitoring the rates based on the different calls’ outcomes. A p-chart for each rate is plotted every day. P-charts have the interviewers on the horizontal axes and the values of the rate on the vertical axes. The control limits are calculated supposing a binomial distribution for the rate. We are also using a non-parametric Shewhart chart for multivariate process control. We identify a situation under statistical control using the data of the previous edition of the survey: we turn the variable of the first edition (the sums of each possible call’s outcome per interviewer) into factorial co-ordinates, through the Binary Correspondence Analysis, and we use the tool of the nested convex hulls in order to obtain data-driven partitions of the sample space. From these partitions we are able to estimate the p-values. Finally, we obtain the position of the new observations on the factorial axis and their empirically p-values.