Tests for Homogeneity of Ordered Co-Gaussian Modes

• Activity Number: 188
• Type: Contributed
• Date/Time: Monday, July 30, 2012 : 10:30 AM to 12:20 PM
• Sponsor: General Methodology
• Author(s): Xiao Zhang*+ and Michael Paul McDermott and Govind Mudholkar
• Companies: University of Rochester and University of Rochester and University of Rochester
• Keywords: Order-restricted ; Likelihood ratio test ; Combining p-values ; Uniformly most powerful

Abstract

Data from unimodal, right-skewed distributions with positive support arise frequently in practice. The Co-Gaussian distribution, a reparameterized ver- sion of the root-reciprocal inverse Gaussian (RRIG) distribution, is a con- venient model for such data. This distribution is indexed by a parameter that is the mode of the distribution. We develop and evaluate likelihood ratio tests for the null hypothesis of equality of ordered modes from several Co-Gaussian distributions. We also develop tests for homogeneity of ordered Co-Gaussian modes based on combining independent P-values. The approach involves decomposing the overall null hypothesis into several nested component hypotheses, each of which can be tested using a uniformly most powerful unbiased size a test. The resulting p-values are independent under the null hypothesis and can be combined using Fisher's method to obtain an overall test.