Statistics and Biological Research

Abstract

IN a letter to NATURE of Aug. 17, Dr. R. A. Fisher has given a clear statement of his reasons for believing that the methods set out in his “Statistical Methods for Research Workers” are adequate to deal with all problems arising in biological research as ordinarily conducted. Those statisticians who, like “Student”, feel that there is need for the study of non-normal distributions particularly in connexion with small samples, might perhaps state their position as follows. In the first place, from their own experience they do not feel so confident as Dr. Fisher that the normal distribution and tests based upon it are adequate in all forms of biological work; in the second place, they are concerned with many non-biological problems in which forms of variation occur ranging from very slight to extreme degrees of a-normality. It is not, therefore, an academic but a very real problem for them to obtain some more exact appreciation than seems hitherto to have been reached, of the point at which a-normality of distribution renders the tests of ‘normal theory’ inaccurate or inefficient.