Generalized regression estimation for a two-phase sample of tax records / John Armstrong and Hélène St-Jean. : CS11-0011/1993E-PDF

"Data from tax returns are used to obtain estimates of Canadian economic production for small businesses. Most variables of interest are not available in machine-readable form and must be obtained from source documents. Sampling is needed to avoid prohibitive data capture costs. Estimates are required for domains defined using four digits of Standard Industrial Classification (SIC) code but only the first two digits of SIC can be reliably obtained from administrative sources. A two-phase sample design using Bernoulli selection is employed. The estimation methodology involves adjustment of Horvitz-Thompson weights to compensate for differences between actual and expected sample sizes within poststrata. This approach does not use all available information. Some economic variables, particularly gross revenue, are available in machine-readable form for all tax records. The possibility of FA improving estimates using this information is examined in the context of generalized regression estimation"--Abstract.

Lien permanent pour cette publication :
publications.gc.ca/pub?id=9.846200&sl=1

• Format MARC XML
• Format MARC HTML

Demander des formats alternatifs