28 



BULLETIN 1004, U. S. DEPARTMENT OF AGRICULTURE. 



for these years. It may be either above or below the point that might 

 be established if the results of a larger number of years were avail- 

 able. 



The accuracy of the linear-regression equation in expressing the 

 relation between water use and yield- at this station is shown in 

 Table 5. This table gives the total water used in inches, the actual 

 yield in bushels per acre, and the yield in bushels per acre as cal- 

 culated from the water use by the linear-regression equation. 



Table 5. — Actual and computed yields of wheat per acre on pjat A at Edgeley, 

 N. Dak., in the 10 years stated. 



[The quantity of water used during the growing season as stated in the table forms the basis of the computed 

 yield as calculated by the linear-regression equation M=8.16-t-0.20 e.] 





Water 

 used. 



Yield. 



Year. 



Water 

 used. 



Yield. 



Year. 



Actual. 



Com- 

 puted. 



Actual. 



Com- 

 puted. 



1907 



Inches. 



9.50 

 10.85 

 13.03 



6.41 

 10.50 



Bushels. 



4.1 



13.3 



28.3 



4.0 



.8 



Bushels. 



6.7 

 13.5 

 24.4 





 11.7 



1912 



Inches. 

 15.77 

 11.29 

 11.04 

 15.28 

 9.10 



Bushels. 

 35.0 

 16.3 

 11.3 

 34.8 

 8.3 



Bushels. 

 38.1 



1908 



1913 



15.7 



1909 



1914 



14.4 



1910... 



1915... 



35.6 



1911 



1917 



4.7 









Table 5 shows how accurately in this case the yield of wheat can be 

 determined from the quantity of water used. In only two years, 1910 

 and 1911, did the computed yield differ greatly from the actual yield. 

 In 1910 a yield of 4 bushels per acre resulted from a use of water con- 

 siderably below the 8.16 inches theoretically supposed to show a 

 zero yield, and in 1911 a yield of only 0.8 bushel per acre was secured 

 with a total use of 10.5 inches of water. In 1911 hot winds badly 

 damaged the crop and seriously reduced the yield. This represents 

 a year in which climatic conditions overbalanced the quantity of 

 water used in determining the yield. The closeness of the computed 

 yield to the actual yield at this station is probably due in part to the 

 fact that at Edgeley crops feed to a depth of only about 2 feet. The 

 inclusion of only 2 feet of soil no doubt makes the size of the experi- 

 mental error in the determination of water used at the station small. 



At North Platte the data of nine years are available. The correla- 

 tion between the total water use and yield is 0.85 ±0.06. The year 1911 

 is eliminated because of the zero yield. This failure was the result 

 of a use of only 1.47 inches of water. Since there might still have 

 been a failure had more water been used, this year does not establish 

 any definite point, and consequently is not included in making the 

 correlation or in determining the linear-regression equation. 



The linear-regression equation determined is M=6.22-|-0.40 e. 

 Figure 10 shows the points established by water use and yield in 

 the several years, together with the line established by the linear- 

 regression equation. 



At North Platte the points established by the water use and yield 

 each year do not lie as close to the line established by the linear- 

 regression equation as those at Edgeley. They do, however, group 

 themselves near enough to the line to indicate that it is not far wrong. 



