|
Activity Number:
|
545
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 2, 2007 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistical Computing
|
| Abstract - #309286 |
|
Title:
|
A Realization of the Functionalized General Trichotomy in Logistic Regression
|
|
Author(s):
|
Ligong Chen*+ and Yongmei Chen and Jennifer Cullen
|
|
Companies:
|
Center for Prostate Disease Research and Center for Prostate Disease Research and Center for Prostate Disease Research
|
|
Address:
|
1530 East Jefferson St, Rockville, MD, 20852,
|
|
Keywords:
|
Functionalized General Trichotomy ; Piecewise Regression ; Logistic Regression ; Double Thresholds ; Convergence Rate of Residuals ; Continuity Test
|
|
Abstract:
|
Based on symmetrically or asymmetrically iterative segmentation, we introduced a functionalized general trichotomy (FGT) for regression analyses. Convergence rates (CR) of means of combined absolute residuals were used as weights per iteration to directly estimate weighted means and confidence intervals of two thresholds. An asymptotic Chi-square statistic was constructed to test differences between the full and segmented models, as well as a statistic to test continuity, and a statistic to evaluate goodness of fit. A quadratic curve of the CR was identified and confirms that the thresholds corresponding to the maximal CR are biased and not optimal estimates for a population. The Bootstrap was also deemed to be of limited use for this method. A study sample obtained from a large public use dataset is analyzed with logistic regression to illustrate the computation and methodology.
|
