232 



BIOLOGICAL EFFECTS OF RADIATION 



probability that the variation in the results over this range of intensities 

 is nothing more than would be expected by chance in sampling from the 

 same universe, and it would therefore be dangerous to elaborate a 

 scientific theory to explain this variation. 



It should be emphasized that significance tests are but preUminary 

 steps and do not constitute an analysis of the problem. If we find that 



Table 4. — Distribution of Observed Mortality Rates Compared with a 



Chance Distribution 



an observed difference is not significant, we shall attempt no explanation 

 of this difference ; but if the observed difference is significant, we proceed 

 to the examination of the possible hypotheses which might account for 

 the observed facts. 



TREATMENT OF MEASUREMENTS 



The preceding section dealt with problems in which the character 

 under observation was one that had a two-way classification, such as 

 died or did not die, but in many cases we are interested in a particular 

 factor of the biological form that is subject to measurement on some 

 scale. In such cases we have a measurement for each individual organism 

 of an experimental and control group and the statistical problem is that 

 of summarizing the measurements in such a way that the two groups 

 may be readily compared. We may, if we choose, select some particular 

 point on the scale and state for the experimental and the control group 

 the percentage of animals that fall above and below this point, and then 

 compare these percentages by the method of the preceding section. In 

 so doing, however, we lose most of the advantage of the measurement, 

 for the treatment proceeds as though the measurements had been taken 

 on a scale having but one division point. A better treatment of such 

 measurements gives us some notion of the point on the scale at which the 

 measurements center and some idea of the extent to which the individuals 

 vary from this centering position. 



While there are an infinite number of centering points that may be 

 defined, there are but three that are in common use, the arithmetic mean, 



