|
Activity Number:
|
284
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, July 31, 2007 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Biometrics Section
|
| Abstract - #308236 |
|
Title:
|
Precision of Double Sampling with Error-Prone and Error-Free Measures and Optimum Case-Control Study Design
|
|
Author(s):
|
Jun-mo Nam*+
|
|
Companies:
|
National Cancer Institute
|
|
Address:
|
6120 Executive Blvd, EPS / Room 8028, Rockville, MD, 20892,
|
|
Keywords:
|
Double sampling ; Efficiency ; Fallible classification ; Gold standard ; Optimum sample size determination ; Validation sub-study
|
|
Abstract:
|
We consider double sampling in which cases and controls are classified as exposed or not-exposed by an inexpensive but error-prone measurement in the primary sampling and a random sub-sampling is classified additionally using the expensive error-free measurement. We present the optimum sample sizes which maximize precision of a true odds ratio estimator for a fixed total cost and also those sample sizes minimizing a total cost for a given value of precision. Examples are given using a sudden infant death syndrome and maternal use of antibiotics study and an asbestos-exposure and lung cancer study. A gain in precision and a cost-saving resulted from the optimum design are substantial when a relative cost of the gold standard is high. The optimum double sampling is also compared with a single sampling where subjects are classified by the gold standard exclusively for a fixed total cost.
|
