MORTALITY STUDIES IN R. I. REDS 9 



standard error around this line was calculated. The data actually fitted a 

 straight line since a dispersion of twice the value of the standard error included 

 more than 95 percent of the actual values. On this basis a consistent upward 

 or downward trend in mortality rates for the experimental period could be noted. 



The significance of the rate of change was tested by the method of Van Uvfen 

 suggested by Pearl (1940). The standard error of the slope was calculated to 

 test its significance. Fisher's t test cannot be used to compare mortality rates 

 in the different lines, because this test gives no consideration to the element of 

 time, which is vitally important in all experiments where different lines are 

 developed through a period of jears. 



The control line showed an increase in mortality rate during the experimental 

 period from a calculated initial value of .90 percent in 1935 to a calculated final 

 value of 6.36 percent in 1942. This line had a positive slope of .78 + .148. This 

 increase is statistically significant because the slope is more than five times its 

 standard error. The high mortality line also showed a very significant increase 

 in mortality rate. The initial calculated rate was —.25 percent and the final 

 calculated rate was 7.17 percent. The rate of increase was 1.06 + .136 percent. 

 The line bred for low mortality did not behave consistently during the experimen- 

 tal period. There was a calculated initial mortality of 8.42 percent and a final 

 value of 3.73 percent. The rate of change measured by the slope was — .67 ± .456. 

 There was definitely no significant decline in mortality rate in this line. In the 

 miscellaneous group an increase in mortality rate was observed. There was a 

 calculated rate of 2.03 percent in 1935 compared with a calculated final rate of 

 5.88 percent. The rate of increase was represented by a slope of .55 ±.264. 

 The odds are about 20 to 1 that this increase was significant. 



In general, these data indicate that there was a tendency for the mortality 

 rates to increase in three lines during this experimental period. Although the 

 mortality rate in the low line did not increase, neither did it exhibit a significant 

 decline. In other words, there was no significant change in mortality rate in 

 the line selected for low mortality. It seems probable, therefore, that selective 

 breeding was effective in preventing any increase in mortality rate in this line 

 despite the abnormally high rate shown in 1938. 



Comparison of Changes in Mortality Rates 



Since the rate of change in mortality rates has been calculated for each line 

 between the ages of 8 weeks and 6 months, it is possible to determine whether 

 the rates of change in the lines have been different through the experimental 

 period. The difference in the rate of change between the high and low mortality 

 lines is of greatest interest. In this instance the difference between the slopes 

 of the two lines was 1.06 — ( — .67) or 1.73 ± .48. Since the difference is more 

 than three times the standard error of difference, it appears that the two lines 

 do differ significantly in their rate of change in mortality rate. 



A comparison of the high mortality line with the control line gave 1.06 — .78 

 or .28 ± .20. In this instance the difference was scarcely greater than the standard 

 error of difference and was clearly not significant. 



When the low mortality line was compared with the control line, the values 

 were .78 — ( — .67) or 1.45 + .48. In this comparison the difference is three 

 times the standard error of difference and is therefore significant. It is logical 

 to assume that the low mortality line differed significantly in mortality rate 

 from both the high mortality line and the control line. 



