Estimating the odds ratio under double sampling /
| Main Author: | |
|---|---|
| Other Authors: | , , |
| Format: | Thesis Book |
| Language: | English |
| Published: |
1991.
|
| Subjects: | |
| Online Access: | Link to OAKTrust copy |
| Abstract: | Increasingly, validation substudies are incorporated as part of epidemiologic designs. In the initial stage, a sample of individuals is classified by disease outcome and exposure status using a fallible classifier only. In the second stage, a subsample of individuals is classified further using a "gold standard" exposure classifier. A sampling scheme of this type is known as a double sampling plan. Our primary interest in this study is to estimate the odds ratio, ψ. The odds ratio is a measure of association between the disease level, D and the true exposure level, X. Two different methods that are commonly used to estimate ψ are maximum likelihood and the "matrix" methods. We compare the performance of these two estimation methods by comparing variances and biases of the estimators of the odds ratio. We treat case-control and cohort studies separately. Under the condition of nondifferentiability, the likelihood criterion equations do not appear to yield analytical solutions. Thus, in addition to the maximum likelihood estimators, we investigate pseudo-maximum likelihood estimators of the odds ratio for case-control and cohort studies under the assumption of nondifferential misclassification. We use the delta method to compute the variances of the asymptotic distributions of the estimators of the odds ratio. Further, a simulation study numerically evaluates variances and biases under various combinations of true parameters. |
|---|---|
| Item Description: | Typescript (photocopy). Vita. "Major subject: Statistics." |
| Physical Description: | xii, 110 leaves : illustrations ; 29 cm |
| Bibliography: | Includes bibliographical references. |