|
Activity Number:
|
275
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 5, 2008 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Biometrics Section
|
| Abstract - #301039 |
|
Title:
|
Optimum Two-Stage Sampling with Inexpensive Error-Prone and Expensive Error-Free Measures for Designing a Case-Control Study
|
|
Author(s):
|
Jun-mo Nam*+
|
|
Companies:
|
National Cancer Institute
|
|
Address:
|
Executive Plaza South/ Romm 8028, Rockville, MD, 20852-7244,
|
|
Keywords:
|
cost efficient double sampling ; fallible classification ; gold standard ; optimum sample size ; validation sub-sampling
|
|
Abstract:
|
We consider a case-control study design using two-stage sampling in which case and controls are classified as exposed or not-exposed by an inexpensive but error-prone measure in the 1st stage and it's sub-sampling is classified by the expensive but error-free measure in the 2nd stage. We provide the optimum proportion for validation size to the primary one which maximize precision of estimation of a true odds ratio for a given total cost, and present the optimum sample size formulae for designing a case-control study. We compare our optimal design with McNamee's optimal designs. Examples are given for numerical illustrations. A gain in precision of estimation and cost reduction resulted from the optimal allocation in a comparison with those not using the optimal rule is substantial when a unit cost ratio of the gold standard to a fallible classification is very high.
|
