VI 



STATISTICAL TREATMENT OF BIOLOGICAL PROBLEMS IN 



IRRADIATION 



Lowell J. Reed 



Department of Bio statistics, Johns Hopkins University, Baltimore 



Percentage comparisons. Treatment of measurements. Treatment of lethal dose. 

 Curve fitting. References. 



Biological problems have a degree of variability such that when 

 treated quantitatively they become automatically statistical in character, 

 and when we add to this degree of variability that inherent in the field of 

 radiation, it becomes doubly necessary to treat the observations obtained 

 with appropriate statistical technique. The problems encountered in 

 this field involve no processes of quantitative logic not discussed in some 

 of the various textbooks on statistical method and it would be inappro- 

 priate in this place to attempt to cover the ground usually elaborated by 

 such books. It may, however, be of value to consider certain types of 

 problems that occur reasonably often in the study of the biological 

 effects of radiation and to present methods for analyzing them statisti- 

 cally. Since space prevents a detailed discussion of all the points 

 involved, the reader will be referred for certain routine treatments to 

 such standard textbooks as those of Yule (10), Pearl (5), and Tippett (9). 

 The emphasis of the present discussion is placed rather on the interpreta- 

 tion of the statistical constants than on their derivation, and although it 

 cannot hope to train workers in this field in statistical method, it aims to 

 give direction to their studies. 



PERCENTAGE COMPARISONS 



In the study of the effect of irradiation on biological forms, the 

 simplest statistical problem arises when one group of animals has been 

 subjected to radiation of some type, a similar group of animals has not 

 been subjected to radiation, and the observed end result is of the yes or 

 no type, that is, the animals died or survived, the eggs hatched or did not 

 hatch, etc. Such problems lead to a comparison of the percentages of 

 animals that reacted in the experimental and control groups. Compari- 

 son of these percentages is so simple and direct that the only statistical 

 question involved is that of the degree of reliability of the observed 



227 



