296 HARRIS, KIRKPATRICK, BLAKESLEE, WARNER AND CARD 



Turning now to the problem of the comparison of periods of one month 

 and of two months as bases of prediction, and testing the efficiency of 

 these two periods on the egg production of comparable remaining periods 

 of time, we note that the differences in the two final columns of tables 5 

 to 7, expressed either in number of eggs or in percentages of the total 

 production, are small. Thus the differences for the average deviation 

 with regard to sign are all less than 3 eggs and all less than 6 percent. 

 Most of the differences are far smaller than this. In some cases the predic 

 tion from a single month gives the better result; in others prediction from 

 two months gives the better result. The differences in the errors without 

 regard to sign as obtained by the two methods are even smaller. No 

 difference amounts to as much as a single egg per year. The large differ 

 ences in the percentage errors by the two methods are found exclusively 

 in the later months of the year where the total production is low. Com 

 parable, but numerically somewhat different, results are found for the 

 square root of mean square deviation. 



Thus it is clear that there is little practical difference between single- 

 month and two-months production as bases of the prediction of the egg 

 record of a subsequent period. 



Prediction of annual production from the sum of three monthly records 



The equations required for the prediction of annual production from 

 the combined record of three consecutive months are the following: 



Months from which prediction is made Prediction equation 



November, December and January E = +126.742 + 1.770 (e\ + e 2 + e 3 ) 



December, January and February E = +113.940 + 1 .951 (e 2 + e 3 + e 4 ) 



January, February and March E = +82 . 129 + 2 .266 (e 3 + d + e 6 ) 



February, March and April E = +50.502 + 2.323 (e 4 + e b + e 6 ) 



March, April and May E = +29.450 + 2 .267 (e s + e 6 + 



April, May and June E = + 19 . 349 + 2 . 324 (e 6 + e 7 + 



May, June and July E = +23 . 786 + 2 . 233 fa + e s + 



June, July and August E = +41 .079 + 2.065 (e + e 9 + 



July, August and September E = +67.078 + 1 .895 (e 9 + do + 



August, September and October E = +97 . 699 + 1 . 794 (e i0 + e a + 





The second section of table 8 shows the average deviation with regard 

 to sign of the annual egg production predicted from the combined record 

 of 3 consecutive months from the performance of the 415 White Leghorn 

 birds studied at Storrs in 1917-1918. 



The results show that the trimonthly totals, like the monthly records 

 and bimonthly totals considered in preceding sections, give excellent predic 

 tions. December to February, January to March, March to May, and 



