Use mean score, IPW, and MI methods to estimate the sensitivity and specificity of the test in Example 11.1.

Transcribed Image Text:

Example 11.1 Suppose that we are interested in estimating the prevalence of a disease based on a sample of 1000 subjects randomly selected from the population of interest. Summarized in Table 11.1(a) is the information about the number of diseased subjects and test results from a screening test. Now suppose that 50% of those tested negative have their test results confirmed, with the results shown in Table 11.1(b). TABLE 11.1: A hypothetical study of a diagnostic test. Test Diseased Positive Negative Test Diseased Positive Negative Yes 450 50 Yes 450 25 No 100 400 No 100 200 (a) Complete Data (b) Verified Subsample === If the remaining subjects without their negative tests validated are deleted, the naive estimate of prevalence based on the subset in Table 11.1(b) is 475 = 61.3%, much higher than 50% obtained based on the original complete data in Table 11.1(a). []