George W. Snedecor

Correlation and Machine Calculation

males in the RLL and LRIR groups and that seine curious discrepancies occur in the percentages for the sexes, which, however, can not be adequately stated on the basis of the figures given in the table. In comparison of my figures w-ith those of Lutz, it is interestinTg to note that his -unclassified group gives a higher percentage of right thumnbs uppermost for both sexes than my figures show. There are several conjectural explanations for this difference. It may be due to the fact that his groups were smaller than mine. On the other hand, it is possible that his groups mere mnore racially homiogenous and more strongly right-handed than myv groups. It may be that variation in the frequency with which dextrality types occur is a racial characteristic of some im-1portance. Lutzs percentages -were apparently obtained fromt Scotch subjects. Whether or not the Scotch are a, strongly rioht-bland(ed people, either by inheritance or habit, remains to be determined. JUNE E. DOWNEY UNIVERSITY OF WYOMING

The Improvement of Statistical Techniques in Biology

THE ENTRANCE of the statistician into the field of experimental design is the most outstanding feature of modern statistics. No more than a decade past, the statistician was distinctly on the defensive. When the data from an experiment were placed on his desk, he was supposed to produce some method of treatment irrespective of design. In some instances he was lucky; in others, baffled. Under the leadership of R. A. Fisher, the statistician has become the aggressor. He has found that the key to the problem is the intimate relation between the statistical method and the experimental plan. The latter must be capable of producing not only an unbiased estimate of the desired effects, but also a valid estimate of error. Many of the plans in use did neither. The statistician is now able to assert that properly designed experiments yield unqualified answers to the questions asked. It is incumbent upon the investigator to plan his experiment in accord with established principles, if he wishes to avail himself of adequate statistical methods. Only if the plan and the method are compatible may unambiguous conclusions be drawn from the statistics. To insure this, the statistical method must be decided upon along with the experimental plan before the laboratory or field work is initiated. The integration of plan and method lead to efficient designs.

On a Unique Feature of Statistics

Abstract * Presidential address delivered at the 108th Annual Meeting of the American Statistical Association on December 28, 1948.

Biostatistical Problems Involved in the Standardization of Liquid Household Insecticides

INTERSTATE commerce in insecticides, drugs and other chemicals that are intended to affect living organisms favorably or adversely is controlled by the Food and Drug Administration of the U. S. Department of Agriculture. The laws require that contents and performance of such products be truthfully stated on the label. Some products are not susceptible to accurate chemical analysis. Others are such that chemical analysis cannot be interpreted in terms of biological performance. In order to check on the performance of such products it is therefore necessary to test them against living organisms and to estimate their value by the effects produced. These so-called biological tests are made not only by law enforcement agencies but by reputable manufacturers who wish to maintain the quality of their products. Biological tests are most necessary for the determination of the quality of liquid insecticides that are used for the control of household insect pests, such as houseflies, mosquitoes, bedbugs, cockroaches, etc. These insecticides consist of a more or less refined kerosene oil in which are dissolved materials extracted from certain plants, or certain synthetic organic compounds, or combinations of natural and synthetic compounds. To these materials a perfume may be added. The finished product is a dilute solution in oil of complex organic compounds which are difficult or impossible to evaluate chemically. The value of such liquids is determined by testing them against one of the species of insects that they are intended to control. The method now used almost exclusively for this purpose is the outgrowth of a method published by Peet and Grady in 1928.1 Having learned how to rear houseffies in large numbers in the laboratory throughout the year, they set up a 6 ft. cubical testing chamber within which houseflies were released and into which an insecticide was then sprayed, filling the chamber with a mist of oil. As the chamber was large enough to permit normal flight of the insects, practical conditions of application of the insecticide in a room were simulated and at the

On the Design of Sampling Investigations

Abstract Professor Snedecors discussion of the design of sampling investigations was prepared early last fall as a speech which was delivered at meetings of the Chicago and Central Indiana Chapters of the American Statistical Association. No basic revisions have been made since the 1948 presidential elections.