We have two sample covariance matricea of Bize p x p , where p is the number of variables denoting measurements of arterial blood pressnre reoorded in varions pOeitiom and method6 for women and men. The h y p o t h e on equality of the two ooverianoe matrices waa rejected and we went to find, why thin happened and in which covarianoes the two matricea differ.
In 0 2 we explore the primary (observed) data looking for outliers and deviations from normality. We do it by drawing
In 8 3 we look for common principal components. W e use here Flury's method. We find that the data for blood pressure systolic might have a common principal axes aptem (P=0.0866), while for blood pressure diastolic this is dubious (p ==0.0112). The varianow of the first prinoipal component found for woman am a little larger than those for men, and this faot explah the rejection of the cleseical teet &=& on the equality of the two covarianoe matrica.
After applying a common rotation to the two cova~anoe matrices we reduoe them fo a nearly die-gone1 form. Looking a t the non diagonal elements we me which oovarianoea do not fit inta the model.
I n 0 4 we repeat the calculations from 8 3 for covariance matrices calculated from the same data using robust methods.
plots.
Key WMdS: Covariance structure ; Common principal components ; Simultaneous diagonalisation of two positive definite covariance matricee ; Reproduced covariances.
I. I n t d u c t i o n to the Medical Problem
Traditionally arterial blood preasure (b.p.) is recorded using a mercurial sfigmanometer from which the indications of the blood pressure are read by the physician (nurse). In Lower Sileaia Medical Diagnosia Center in Wrodaw the measurements are recorded automatically by the Avionica 1900 devim connected directly with computer core. To compare these two methods of r e o o w the blood pressure an experiment was Conducted during which for the =me patient in a randomized mquenm measurements by both methods were re-corded. The measurements were done in 3 positions of the patient : standing, sitting, lying. Finally for each patient we got p='6 measurements, which in the forthcoming presentation will be called 60 Blom. J. DO (1987) 8
' .