A Ranking Test in the Biological Sciences

Laboratory analyses of biological materials are ranked in order of magnitude and summed across materials to give a list of laboratory scores. Under the assumed hypothesis that there is in fact no difference between laboratories, Monte-Carlo techniques are used to establish two-tailed 5% rejection limits for various combinations of laboratories and materials. The hypothesis that there is no difference between laboratories is rejected if any laboratory\u27s score lies outside the 5% limits. Suppose that one needs to run a group of tests on a particular set of materials (chemical or biological), using a number of different laboratories, and wishes to insure before starting that the laboratories are reliable, i.e., that (a) they run the test according to required specifications or directions and (b) if they run the same test twice, they will get, within some tolerated instrument variation, the same results. I shall develop a statistical test here based on the ranked laboratory results which does not assume that the data have any particular distribution. The basis for this work was done by Dr. W. J. Youden of the National Bureau of Standards (1963), but his work is done primarily with a view to industrial applications. I have endeavored here to simplify the statistical procedures and to stress biological applications by way of examples