|
Activity Number:
|
160
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 9, 2004 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #301093 |
|
Title:
|
Why Large Design Effects Can Occur in Complex Sample Designs: Examples from the NHANES 1999-2000 Survey
|
|
Author(s):
|
David A. Lacher*+ and Lester R. Curtin and Jeffery P. Hughes
|
|
Companies:
|
National Center for Health Statistics and Centers for Disease Control and Prevention and National Center for Health Statistics
|
|
Address:
|
3311 Toledo Rd., Hyattsville, MD, 20782,
|
|
Keywords:
|
design effect ; complex sample ; variance estimation ; NHANES ; laboratory test
|
|
Abstract:
|
The design effect is defined as the ratio of the "true" sampling variance to the hypothetical simple random sample variance assuming the same point estimate. Design effects are used to compare alternative designs to determine the effective sample size for analysis and to adjust confidence intervals for estimates based on complex designs. Design effects have been examined for the National Health and Nutrition Examination Survey (NHANES) for 1999-2000. Although design effects for most estimates based on NHANES were less than three, design effects for means of laboratory tests were much higher, ranging up to 40. Design effects of means were larger for laboratory tests with small between-person coefficients of variation. It was determined that small systematic shifts in laboratory test values apparently led to larger between-sampling unit component of sampling error and increased the design effect. Simulations were done to compare the effect of analytical bias and the effect of precision of laboratory methods on the design effect of the mean. For the NHANES data, the bias of the laboratory method influenced the design effects of means more than the precision of the laboratory test.
|
