Applications of variance due to imputation in the Survey of Employment, Payrolls and Hours / Pierre Felx and Eric Rancourt. : CS11-617/01-9E-PDF

"Imputation is a well-known method for dealing with nonresponse in periodic business surveys conducted on a monthly basis. It is also well known that the ordinary variance formula underestimates the variance when imputation is used. As a direct result, the amount of variation due to imputation becomes a necessary calculation. This is especially important when estimates are required for arbitrary domains. Methods have been proposed for calculating the amount of variation due to imputation for arbitrary domains, for many different types of imputation, for a simple random sample without replacement. In this paper, we study the model-assisted approach for stratified simple random sampling when mean, trend and ratio imputation is used. The results obtained are applied to the Canadian Survey of Employment, Payrolls and Hours to better estimate the total variance, and to help determine whether there may be breaks between two redesign phases of the survey"--Abstract.

Lien permanent pour cette publication :
publications.gc.ca/pub?id=9.840391&sl=1

• Format MARC XML
• Format MARC HTML

Demander des formats alternatifs