An evaluation of statistical matching methods / by J. Armstrong. : CS11-617/90-3E-PDF

"Statistical matching is a technique for combining information from two microdata sets that does not require the presence of unique identifiers. Validity of results obtained using statistical matching depends on assumptions about relationships between variables that are unique to each input file. Univariate and bivariate distributions on files created by statistical matching may be subject to additional distortions. Performances of a number of statistical matching methods are examined using synthetic data as well as a inicroeconomic data file created by exact matching of information from the Survey of Consumer Finance and Revenue Canada's tax file. Statistical matching results are compared to those obtained using imputation methods that require a small microdata set containing information about all variables"--Abstract.

Lien permanent pour cette publication :
publications.gc.ca/pub?id=9.840597&sl=1

• Format MARC XML
• Format MARC HTML

Demander des formats alternatifs