Quantile estimation under successive sampling (Q2488409)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Quantile estimation under successive sampling |
scientific article |
Statements
Quantile estimation under successive sampling (English)
0 references
24 May 2006
0 references
For a finite population \(U\) some variable is observed on two successive occasions (current and previous). The sample drawn on the previous occasion is \(X\), and the sample on the current occasion is \(Y=Y_m\cup Y_u\), where the data \(Y_m\) are obtained from the members of \(U\) which were observed at the previous occasion (matched sample), whence \(Y_u\) are data on previously non-observed members. The \(\beta\)-th quantile of the distribution at the current occasion is estimated by \[ \tilde Q(\beta)=w\hat Q_{Y_m}^H(\beta)+(1-w)\hat Q_{Y_u}(\beta), \] where \(\hat Q_{Y_u}(\beta)\) denotes the empirical quantile of \(Y_u\) and \[ \hat Q_{Y_m}^H(\beta)=H(\hat Q_{Y_m}(\beta),\hat Q_{X_m}(\beta)/\hat Q_{X}(\beta)), \] \(H\) can be some appropriate function, e.g., \(H(q,t)=q/t\). The asymptotic normality of the estimators is stated and their asymptotic variances are evaluated. The optimal choice of \(w\) is discussed. Estimators for different \(H\) are compared at a real data example.
0 references
finite population quantile
0 references
matching fraction
0 references
asymptotic variance
0 references