TRAINING KEY 205: Model-Assisted Estimation for Domains

Practical Methods for Design and Analysis of Complex Surveys.
Risto Lehtonen and Erkki Pahkinen

TRAINING KEY 205: Model-Assisted Estimation for Domains



INTRODUCTION

Example 6.2. Estimation of domain totals by design-based model-assisted methods under SRSWOR sampling. We illustrate the domain estimation methodology by selecting an SRSWOR sample (n = 1960 persons) from the OHC Survey data set (N = 7841 persons) and estimating the total number of chronically ill persons in the D = 30 domains constructed. In the population, the sizes of the domains vary with a minimum of 81 persons and a maximum of 517 persons. The domain proportion of chronically ill persons varies from 18 to 39%, and the overall proportion is 29%. The intra-domain correlation of being chronically ill (binary response) and the age (in years) varies from 0.08 to 0.55; the overall correlation is 0.28.



A) MODEL-ASSISTED ESTIMATION FOR DOMAINS

We will demonstrate how to perform model-assisted estimation for domains in the setting of Example 6.2. Further instructions are given once you start.
 

 


B) MODEL-ASSISTED ESTIMATION FOR DOMAINS WITH DIFFERENT SAMPLE SIZES

We will demonstrate how to perform model-assisted estimation for domains with different sample sizes. The Horvitz-Thompson (HT) estimator and the generalized regression (GREG) estimator of a domain total are compared by examining the standard error and coefficient of variation estimates. Further instructions are given once you start.
 

 


C) INTERACTIVE SAS USE

Please download the SAS code for your own further training. Select your own sample or several samples and exercise model-assisted estimation for domains with different sample sizes for a SRSWOR sample. The macro parameters used in the application are n = sample size, seed = seed for the random number generator (default seed=99919481957).

NOTE! You need to have access to SAS in your computer.
 

SAS code (key205-macro.sas)
Data (saem.sas7bdat)