Online Program

Comparing Estimation Methods in Product Surveys
*Ismo Teikari, Statistics Finland 

Keywords: Rotation method, cold deck estimation, PRODCOM statistic

Sometimes we meet a sampling problem to have only one or few occurrences of the phenomenon under study. One extreme example is the PRODCOM statistics. Due to the 5700 PRODCOM codes, there happen to be many headings having only few producers. We have no control over all PRODCOM codes using the traditional random sampling.

Due to the few occurrences we miss some products and overestimate some if we use the sampling method. To avoid this problem, samples are rotated and the cold deck estimation, called rotation method, is used. The last year production value augmented by an index number was used as an estimate.

Three estimation methods, the HT-, the ratio-, and the rotation method’ is used. The sampling method was Simple Random Sampling. The test frame was the PRODCOM statistics 2003.

Some M-C simulations were performed to test the efficiency of three estimators. Some cases, number of occurrences, were studied. The cases were compared using CV of the standard deviation and MSE. HT-estimator was unbiased but rotation method gave the smallest MSE in most cases. It gave smaller MSE than the ratio estimator when occurrences were 36 and smaller MSE than HT when it was 94.