ABSTRACT
<p><b>OBJECTIVE</b>To explore the way for testing of multivariate in therapeutical equivalence.</p><p><b>METHODS</b>By improved Hotelling T(2) test, the mean vector of two groups was tested. The information of each variable could then be obtained by deducting the their correlations. "Clinical Equivalence of Drug A and Drug B for Postmenopausal Bone Looseness" was used in this research. The primary measure of efficacy was regarded as the improvement rate of four variables with predefined limit of equivalence (2%) with the results of univariate test and multivariate test compared.</p><p><b>RESULTS</b>Under P > alpha, all the variables were nonequivalent. Under P </= alpha, it was necessary to filtrate the variables of unequivalent until the residuals were equivalent. The method of multivariate test was a remedy to the shortage of univariate test.</p><p><b>CONCLUSION</b>With the combination of the filtration of nonequivalent variables and the improved Hotelling T(2) test, the test of multivariate could be done. The method is reasonable and practical.</p>