|
Activity Number:
|
440
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 5, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #305429 |
|
Title:
|
Test Coefficients Regression When the Variables Are Not Independent from Each Other
|
|
Author(s):
|
Leann Myers and Adriana C. Dornelles*+
|
|
Companies:
|
Tulane University and Tulane University
|
|
Address:
|
1440 Canal St, New Orleans, LA, 70112,
|
|
Keywords:
|
correlation ; system of simple regression ; equality of regressions ; heterogeneity
|
|
Abstract:
|
A common problem in estimating regression equations is the correlation among variables. In public health research this is also an issue when we have covariates that alone do not tell us all the information that we want, but together can turn the estimates biased. The objective of this study is to test the equality of regressions coefficients when the variables are interdependent. By breaking down a multiple regression into a system of simple regression equations, we can test the heterogeneity of the set of correlated correlation when there is a common depend variable [ Meng, XL & Rosenthal, R. (1992)] and also test the parallelism among the regression lines. The purpose of this study is to find an innovative solution to get unbiased estimators of the system of regression equation cited above when the lines are found to be non-parallel.
|
