Activity Number:
|
621
- Beyond Linear Regression: Nonlinear Association, Quantile Regression and Generalized Linear Models
|
Type:
|
Contributed
|
Date/Time:
|
Thursday, August 1, 2019 : 8:30 AM to 10:20 AM
|
Sponsor:
|
Section on Statistics in Epidemiology
|
Abstract #304865
|
Presentation 1
Presentation 2
|
Title:
|
Estimating Disparities in Breast Cancer Mortality by Race and Ethnicity
|
Author(s):
|
Ronald Gangnon* and Chrstina Hunter Chapman and Jennifer Bird and Amy Trentham-Dietz
|
Companies:
|
University of Wisconsin and University of Michigan and University of Wisconsin and University of Wisconsin
|
Keywords:
|
disparity;
breast cancer;
logistic regression;
generalized additive model;
shrinkage;
penalization
|
Abstract:
|
Assessments of long-term changes in breast cancer disparities by race/ethnicity require estimates of breast cancer mortality (or incidence) by race/ethnicity for past years, but data is only available for many racial/ethnic groups (Asian/Pacific Islander, American Indian/Alaska Native, Hispanic) for relatively recent years. We propose an age-period-cohort (APC) modeling strategy that borrows strength from the overall (all races and ethnicities combined) APC model to stabilize the APC model estimates for racial/ethnic subgroups. Age, period (year of death) and cohort (year of birth) are entered into a generalized additive regression model as additive natural cubic splines. To stabilize estimates for the race- and ethnicity-specific models, differences between the component functions for all races/ethnicities combined and for each racial/ethnic subgroup will be penalized towards no effect. Smoothing parameters for the various splines will be selected using generalized cross-validation with a BIC-like penalty to avoid overfitting. We illustrate the approach by estimating disparities in the proportion of deaths due to breast cancer between racial/ethnic subgroups from 1965-2016.
|
Authors who are presenting talks have a * after their name.