Survey error modelling in the presence of benchmarks / by Zhao-Guo Chen and Ka Ho Wu. : CS11-617/2001-13E-PDF

Data for a socio-economic variable obtained from a repeated survey contain sampling error. Usually estimates of the variance of the error are obtained in the survey process and published; but estimates of the autocorrelation of the error series (equivalently, a fitted time series model) are rarely given. Knowing autocorrelation of the survey error series, some advanced approaches for predicting the variable can work. This paper proposes a method of modelling monthly survey error using annual benchmarks as additional information under a very general model assumption for the variable. The fitted model is thus more data-based than those obtained from secondary analysis (e.g., Scott, Smith and Jones, 1977) where only monthly data are used.

Lien permanent pour cette publication :
publications.gc.ca/pub?id=9.916556&sl=1

• Format MARC XML
• Format MARC HTML

Demander des formats alternatifs