|
Activity Number:
|
511
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 7, 2008 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #302100 |
|
Title:
|
B-Splines and Bootstrapping for Piecewise Logistic Regression in Complex Samples
|
|
Author(s):
|
Scott W. Keith*+ and David B. Allison
|
|
Companies:
|
The University of Alabama at Birmingham and The University of Alabama at Birmingham
|
|
Address:
|
Ryals Public Health Bld, Birmingham, AL, 35294,
|
|
Keywords:
|
free-knot ; B-splines ; bootstrap ; multi-stage samples ; logistic regression ; nonlinear models
|
|
Abstract:
|
Splines are a flexible application for nonlinear modeling. Estimating both the number and locations of knots as free parameters (join points in a "free-knot" spline) can optimize model parsimony. We designed a spline framework that estimates nonlinear relationships between a binary outcome variable and a continuous prognostic variable in the presence of covariates in complex nationally representative samples. We use direct search methods to maximize likelihood equations with piecewise linear free-knot B-splines. Model selection and inference incorporate parametric and nonparametric bootstrap methodology. Parameter estimates are structured for interpretability so that results look similar to a logistic regression. Unlike other nonlinear approaches, our framework handles complex multistage sampled data. Our presentation will include a summary of methodological details and simulations.
|
