Constructing a complex sample with minimal design effect for a survey that yields estimates for small population groups and for multiple geographic strata is a challenge for survey researchers. This is further complicated when there is an overall capped sample size. This was the situation when the current California Health Interview Survey (CHIS) was designed. CHIS is intended to produce estimates statewide and for individual counties and for several racial/ethnic groups within a maximum sample size of 55,000 households.
Population sizes in California's 58 counties differ dramatically - about 1,000 in the smallest to 12 million in Los Angeles - 64% have populations under 250,000. Small, rural counties required a disproportionately large amount of sample to yield meaningful estimates. CHIS is also designed to produce estimates for several racial/ethnic groups, specifically Latino, American Indian, Filipino, Chinese, Vietnamese, Korean, Asian Indian, and Cambodian. Except for Latino, these groups are relatively small in size. Therefore, they also require disproportionately large samples to yield reliable estimates. These groups tend to reside in large urban counties.
The authors will present their different sampling options. The selected option, with an effective sample size reduction in the range of 20%, included 41 geographic strata, 3 over-sampled cities and 5 over-samples of Asian subgroups. This presentation is intended to assist other survey researchers facing the same design challenge in statewide public health surveys with diverse populations and competing objectives.
See www.chis.ucla.eduLearning Objectives: This presentation is intended to assist other survey researchers facing the same design challenge in statewide public health surveys with diverse populations and competing objectives.
Keywords: Survey, Methodology
Presenting author's disclosure statement:
Organization/institution whose products or services will be discussed: California Health Interview Survey, UCLA Center for Health Policy Research
Disclosure not received
Relationship: Not Received.