Loading...
Please wait, while we are loading the content...
Similar Documents
Replication Variance Estimation for Two-phase Samples
| Content Provider | Semantic Scholar |
|---|---|
| Author | Fuller, Wayne A. |
| Copyright Year | 1998 |
| Abstract | The estimation of the variance of the regression estimator for a two- phase sample is considered. Given the covariance matrix of the first-phase control variables, a replication variance estimator that uses only the second-phase sample is developed. The procedure has computational advantages for surveys with second- phase samples that are small relative to the first-phase sample. Two-phase samples are samples in which a vector of characteristics, denoted by X t where t denotes an individual element, is observed on a relatively large sample. An extended vector, denoted by (X t, Y t) is observed on a subsample of the large sample. We denote the first sample by s(1) and the second sample by s(2). Estimation for two-phase samples is discussed in texts such as Cochran (1977), Sarndal, Swensson and Wretman (1992), and Wolter (1985). Different procedures can be used to combine the information from the first and second samples to obtain estimators of functions of (X, Y ). We will concentrate on the regression method. Variance estimation for two-phase samples based on standard Taylor argu- ments is described in the references previously cited. A jackknife procedure has been suggested for the ratio estimator by Rao and Sitter (1995) and for the regression estimator by Sitter (1997). These procedures involve creating jack- knife replicates using the entire first-phase sample. Sarndal and Swensson (1987) discuss variance estimation for regression estimation in the general two-phase situation. Kott (1990, 1995), discussed variance estimation in a situation where the stratified design for the second-phase differs from the stratified design for the first-phase. He suggests a jackknife estimator of variance for a particular two-phase estimator. Breidt and Fuller (1993) suggested a replication variance estimator for multiphase samples.The Breidt and Fuller procedure is particularly applicable if the first-phase and second-phase primary sampling units are the |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | http://www3.stat.sinica.edu.tw/statistica/oldpdf/A8n48.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
