15 



on one cow was 22 pounds too high, and on another 20.2 pounds too 

 little. It seems probable that in a group of cows used for experimental 

 purposes that errors will about counterbalance each other, and that the 

 results calculated in the two-week tests are not far from the actual milk 

 produced by the cows in a group, though the results may be considerably 

 too high or too low for individual cows, in a short test. 



Error during Four Months. Four cows in the herd were milked 

 most of the time during July, August, September and October by hand. 

 The cow, "Dreamy," was not purchased until about the middle of Sep- 

 tember, hence her record is available for comparison for only about six 

 weeks. "Adelaide" was milked with the machine for about half of July, 

 during which period her milk was weighed every seven days. 



The table shows the error introduced by calculating the monthly milk 

 yield on the basis of four or five weighings monthly. The dates selected 

 were those used for weighing the milk from machine milked cows and 

 multiplying by seven, three or four times and once by three, nine or ten, 

 according to the number of days in the month. An example will illus- 

 trate : During July five weighings were made of the milk given by machine 

 milked cows. The dates were July 2, 10, 17, 24 and 31. Cow No. 98 

 gave 21 pounds on July 2, and was credited with 21x7= 147 pounds 

 for that week ; July 10, 29 x 7 = 203 pounds ; July 17, 27 x 7= 189 pounds ; 

 July 24, 25 x 7=175 pounds; July 31, 26 x 3 = 78 pounds, or a total of 

 792 pounds for the month.. 



During each of the other three months only four weighings were 

 made and the weight of milk given at the fourth weighing was multiplied 

 by 9 or 10 according to the number of days in the month. 



Table Showing Error Introduced by Weekly Weighings Instead ^f Daily, 

 for Four Cows Milked for Considerate Time by Hand and where 

 each Milking was Weighed. 



