Conference Proceeding Article
This paper studies a semi-parametric method for estimating the prevalence of a binary outcome using a two-phase survey. The motivation for a two-phase survey is, due to time, money and ethical considerations, it is impossible to carry out comprehensive evaluation on all subjects in a large random sample of the population. Rather, a relatively inexpensive "screening test" is given to all subjects in the random sample and only individuals more likely to have a positive outcome (cases) will be selected for a further "gold standard" test to verify the outcome. Therefore, individuals with verified outcome form a non-random sample from the population and care must be taken when the data are used for estimating the prevalence of the outcome. This paper proposes a semi-parametric method for estimating the outcome prevalence. It requires only an estimate of the probability of selection into the second phase, given the first phase data. This feature is desirable as in most cases, the probability of selection into the second phase is under the control of the researchers, and even when it is not, can be easily estimated given the data. The proposed method uses the empirical likelihood approach (Owen, 1988), which yields consistent prevalence estimates as long as the probability of selection into the second phase is correctly modeled.
Empiricial likelihood; missing data; surrogate; two-phase sampling
Interfacing modelling and simulation with mathematical and computational sciences: 18th IMACS World Congress, MODSIM09, Cairns, Australia 13-17 July 2009: Proceedings
City or Country
LEUNG, Denis H. Y. and QIN, J.
Semiparametric prevalence estimation from a two-phase survey. (2009). Interfacing modelling and simulation with mathematical and computational sciences: 18th IMACS World Congress, MODSIM09, Cairns, Australia 13-17 July 2009: Proceedings. 1328-1334. Research Collection School Of Economics.
Available at: http://ink.library.smu.edu.sg/soe_research/2088
Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.