Sample sizes for comparing means of two multivariate normal populations

Publisher Website
Google Scholar
Abstract
Sample sixes for the problem of testing the equality of the mean vectors of two nultivariate normal distributions depend on the difference to be detected,. the probability that those differences will be detected, the significance level, and the covariance matrix. The difference to be detected and the variance matrix are embedded in the noncentrality parameter of the noncentral F-distribution.Even though probabilities for the noncentral F-distribution have been tabled in various places these probabilities are not very helpful to the investigator who needs to determine sample sizes. Lauter (1983) presents conservative sample sizes for the two population problem when no information is available about the correlation matrix. The objectivw of this paper is to present additional sample sizes which result when one has certain types of information about the correlation matrix.

This publication has 3 references indexed in Scilit: