With the increase demand of local information, Taiwan is considering using the method of small-area estimation in a Health Interview Survey. Local governments need data in small areas, especially in remote areas in Taiwan, to design suitable programs or policies. The combined synthetic-regression method will be used. Attention will be given to small area with area-specific effects, including age structure and ethnic effects. Data for this study came from the Health Behavior-KAP survey, which was conducted in 2002. The survey sampled 30,000 subjects older than 15 years. Each county was a stratum. Therefore, 23 strata were used. The dataset had the advantage in providing national estimates as well as local estimates. We will identify key variables in predicting features of interests in health behavior. Then the proportion of the local symptom variables to national value will be estimated and used to predict local values. Once an ideal model is established, the local data will be used to validate the estimated value. We hope to establish a model for estimating health related issues, such as disease prevalence and the risk factors in remote area in Taiwan.
