On the Impact of Sampling Error on Modeling Seasonal Time Series

Stuart Scott


The presence of sampling error in an observed time series may obscure underlying features, such as seasonality. Based on simulated series containing sampling error, estimation of the seasonal parameter in ARIMA models is examined, with and without accounting for the sampling error. In the former case, results include cases where the sampling error is incorrectly specified. Sensitivity of estimation to length of the series and relative size of the sampling error is addressed. Empirical results from a large set of employment series are presented and interpreted in light of the simulation results. The series are from the Bureau of Labor Statistics’ Current Employment Statistics survey.