COMMUNICATING VARIANCES FOR INTERPRETATION OF CHANGES AND TURNING POINTS IN REPEATED SURVEYS

The estimates obtained from repeated surveys are used to monitor and analyse changes in variables of interest over time. Major surveys are often conducted on a monthly or quarterly basis and enable identification of important changes in the level or rate of change of variables, including turning points. The samples used by these surveys are often large and involve overalp of the sample over time through use of a sample rotation procedure. The sample size and the correlation structure in the estimates induced by the rotation pattern affect the sampling errors of the estimates and the analysis of changes in them. Users will analyse and interpret the time series of estimates in various ways often involving estimates for several time periods. Despite the large sample sizes and degree of overlap between the sample for some periods the sampling errors can still substantially affect the estimates of movements and functions of them used to interpret the series of estimates. We consider how to account for sampling errors in the interpretation of the estimates from repeated surveys and how to inform the users and analysts of their possible impact