Abstract

A new parametric hypothesis test of equality of two mean intervals for p-dimensional interval-valued (hyper-rectangles) datasets under the assumption of multivariate normality is proposed. We assume the lower bound and the upper bound of an interval are two repeated measurements and then exploit patterned covariance structure to test the equality of two mean intervals. We first employ an orthogonal transformation to obtain two independent hypotheses that is equivalent to the original hypothesis. It can be shown that the two test statistics to test these two independent hypotheses follow Lawley-Hotelling trace distributions with different parameters. The new test is very efficient in small p-dimensional interval-valued sample situations. The performance of the proposed test is illustrated with a real-life example.