On a Kolmogorov-Smirnov type aligned test in linear regression (Q1087269)

The authors consider the regression model \(Y_{ij}=x'_{ij}\beta_ i+\epsilon_{ij}\), \(i=1,2\), \(j=1,...,n_ i\), where \(\{\epsilon_{ij}\}\) are independent random variables with distribution function \(F_ i\), \(i=1,2\). \(x'_{ij}\) is the jth row of a known \(n_ i\times p\) design matrix \(X_ i\). The problem is to test \(H_ 0: F_ 1=F_ 2=F\), F unknown, vs. \(H_ 1: F_ 1\neq F_ 2\) where either (a) \(\beta_ 1\), \(\beta_ 2\) unknown or (b) \(\beta_ 1=\beta_ 2=\beta\) unknown in \({\mathbb{R}}^ p\). Tests of Kolmogorov-Smirnov type based on the residuals are presented. Under some conditions on the matrix \(X_ i\), \(i=1,2\), the tests are shown to be asymptotically distribution free.