第四讲-抽样分布和抽样技术-909402790.pptVIP

For each of the a=10 PSU’s, we obtain the number of children immunized: 9/24, 11/24, 13/24, 15/24, 16/24, 17/24, 18/24, 20/24, 20/24, 21/24 Adding the numerators, there are 160 immunized children The overall proportion immunized is p=160/240 =0.67 Here If the sample had instead been an SRS of n=240 children from all school, then P=160/240 Compared to cluster sampling, the estimated variance of p is considerably smaller for SRS A ratio quantifies the comparison The intra-cluster correlation can be estimated from the design effect: Consider alternative outcome for our sample of a =10 classrooms Homogeneity within, heterogeneity between 0/24, 0/24, 0/24, 16/24, 24/24, 24/24, 24/24, 24/24, 24/24, 24/24 Procedure ----form strata each of size Nh (Wh=Nh/N) ----independent selection n within each strata ----estimate for stratum h, ----then compute an overall estimate Mean: Sampling variance: Design effect: Strata should be internally homogeneous Strata should differ as much as possible from each other Advantage ----Gains in precision ----Administrative convenience ----Guaranteed representation of important domains. ----Acceptability/credibility ----flexibility Population Stratum 1 Stratum 2 Size N 1,000,000 N1 200,000 N2 800,000 Variance S square 1,800,000 S1 square 4,000,000 S2 square 1,000,000 Mean 1400 3000 1000 n1=240, n2=960, what will be v( ) N=1200 what will be For this example, SRS replacement 1.0 b. Cluster: 1+(b-1)roh1.0 c. Stratified: Proportionate f= Sampling distributions Sampling design Sampling techniques Population elements (N) Sample elements （n） Sampling distribution----- Means (or any statistic) distribution from samples (S) Illustration: ----N=6 MMs ----n=2 selected without replacement ----S=15 possible samples 1 2 3 4 5 6 1 1 1 1 1 6 5 4 3 2 3 3 2 2 2 2 4 6 5 4 5 4 4 3 3 6 6 5 6 5 A. Sample realization: When probability selection methods are used, there is not o