Activity Number:
|
315
|
Type:
|
Contributed
|
Date/Time:
|
Tuesday, August 4, 2009 : 10:30 AM to 12:20 PM
|
Sponsor:
|
Section on Statistics in Epidemiology
|
Abstract - #303678 |
Title:
|
On the Confidence Intervals of Attributable Risk using Delta and Bootstrap Methods for a Cross-Sectional Sampling with Confounders
|
Author(s):
|
Tanweer J. Shapla*+ and Khairul Islam
|
Companies:
|
Eastern Michigan University and Wayne State University
|
Address:
|
515 Pray Harrold, Ypsilanti, MI, 48197,
|
Keywords:
|
Attributable risk ; Cross-sectional sampling ; Delta method ; Confidence Intervals ; Bootstrap method
|
Abstract:
|
We revisit the methods of constructing confidence interval estimates of attributable risk using delta method for a cross-sectional sampling design with confounders. We derive a simpler expression for the asymptotic variance of the attributable risk and utilize it in the construction of confidence intervals. However, delta method generally tends to underestimate the standard error, leading to biased confidence intervals in most situations. Therefore, we consider a bootstrap estimate in the construction of confidence intervals. The bootstrap approach is straightforward and appropriate for constructing confidence intervals and it frees from using the relatively complicated expression of the asymptotic variance formula. We justify the performance of the bootstrap approach with an example of the effect of body mass index on cardiovascular disease with race as a confounder.
|