Double Sampling in Estimation of a Ratio.
Abstract
Problems connected with the estimation of the ratio of the means of a finite bivariate population have been considered in this report. The usual estimator of the ratio, based on the means of a bivariate simple random sample drawn without replacement (s.r.s. (w.o.r.)), has been compared with estimators based on alternative double sampling design. Under this design a very large s.r.s. (w.o.r.) is drawn for measuring only one of the variables and a subsample (s.r.r. (w.o.r.)) is drawn out of the first phase units for measuring the other variable. Efficiency and bias comparisons have been made by subjecting each of the competitors to the same budgetary constraint. It turns out that deviation from the usual set-up sometimes leads to better sampling strategies. Additional keywords: Mean square errors; Statistical inference; Multivariate analysis; Approximation(Mathematics). (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 1985
- Accession Number
- ADA160346
Entities
People
- C. Marcotte
- Masab Ahmad
- R. S. Singh
Organizations
- University of Pittsburgh