ABSTRACT
In many quality control settings the product under examination may have two or more related quality characteristics, and the objective of the super vision is to investigate whether all of these characteristics are simultaneously behaving appropriately. In particular, a standard multivariate quality con trol problem is to consider whether an observed vector of measurements x = ( x j , xkY from a particular sample exhibits any evidence of a location shift from a set of “satisfactory” or “standard” mean values ji° = ( jij,..., n°)'. The individual measurements will usually be correlated due to the nature of the problem, so that their covariance matrix E will not be diagonal. In practice, the mean vector |i° and covariance matrix £ may be estimated from an initial large pool of observations.