Cone order association (Q1907826)

The notion of association (A), a type of positive dependence among random variables, was introduced by \textit{J. D. Esary, F. Proschan} and \textit{D. W. Walkup} [Ann. Math. Stat. 38, 1466-1474 (1967; Zbl 0183.21502)]. They developed some of its properties and demonstrated its usefulness in establishing bounds and inequalities in reliability, hypothesis testing, and applied probability. The motivation for defining the concept of cone order association (COA) comes from seeking properties of tests of hypotheses. In particular, order restricted alternatives frequently are such that the parameters of the model lie in a cone. \textit{A. N. Ahmed, R. León}, and \textit{F. Proschan} [Ann. Stat. 9, 168-176 (1981; Zbl 0502.62056)] and \textit{B. H. Lindqvist} [J. Multivariate Anal. 26, No. 2, 111-132 (1988; Zbl 0655.60005), p. 14] have generalized the notion of association in some directions which are considerably different from the extension given here. The type of association we develop here is based on cone orderings as presented by \textit{A. W. Marshall} and \textit{I. Olkin} in ``Inequalities: Theory of majorization and its applications.'' (1979; Zbl 0437.26007). In the next section, we indicate properties of COA and note a relationship between COA and A in some cases where \({\mathcal K}\) is a polyhedral cone. In Section 3 we treat the normal case plus some generalizations to non-normal cases. In Section 4 we exhibit the connection between COA and unbiased testing along with some examples.