scientific article; zbMATH DE number 3999007

From MaRDI portal
Publication:4725518





[For the entire collection see Zbl 0606.00025.] Consider the multiple linear regression model where, for each \(p\times 1\) vector \(\beta \in {\mathbb{R}}^ p\), the observations \(\{(x'_{ni},Y_{ni})\), \(1\leq i\leq n\}\) satisfy \(Y_{ni}=x'_{ni}\beta +e_{ni}\), \(1\leq i\leq n\), with \(x'_{ni}\) denoting the i-th row of the known \(n\times p\) design matrix X, \(1\leq i\leq n\) and where \(e_{n1},...,e_{np}\) are independent random variables (r.v.'s) with respective continuous distribution functions (d.f.'s) \(F_{n1},...,F_{nn}\). \((X'X)^{-1}\) is assumed to exist for all \(n\geq p\). Consider the problem of testing \(H_ 0: F_{ni}(x)\equiv F_ 0(x/\sigma)\), \(x\in {\mathbb{R}}\), \(\sigma >0\), \(1\leq i\leq n\); \(F_ 0\) a known d.f., and the problem of testing \(H_ S: F_{ni}\equiv F\), F a continuous symmetric d.f. around 0, F unknown. All the tests in the paper are based on empirical and weighted empirical residual processes.











This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4725518)