Power Analysis of the Rao-Scott First-Order Adjustment to the Pearson Test for Homogeneity

• Activity Number: 145
• Type: Contributed
• Date/Time: Monday, August 7, 2006 : 10:30 AM to 12:20 PM
• Sponsor: Section on Survey Research Methods
• Author(s): Sunyeong Heo*+
• Companies: Changwon National University
• Keywords: complex sample ; homogeneity test ; Rao-Scott first order adjustment ; Wald test ; distribution of weighted sum of noncentral chisquare random variables ; 2003 NHIS (National Health Interview Survey)

Abstract

In a secondary categorical data analysis of complex sample data, situations often arise when the full variance-covariance matrices of cell proportion estimates are not available but only their variances. In this case, researchers often use the Rao-Scott first order adjustments for inferences. In some circumstances, the type I error rate is greater than the nominal level when using the adjusted test. In addition, there is a concern about the loss of power. I have searched for a way to evaluate powers of the Rao-Scott first order adjustments to Pearson-type test statistics for the homogeneity test, and compared them with the powers of the Wald test. To evaluate power, I used Solomon and Stephens' three-moments approximation. The proposed methods are applied to 2003 NHIS (National Health Interview Survey) public use data.