142 



Bulletin 394 



DISCUSSION OF RESULTS 



A brief discussion of the results furnished by this investigation has 

 been given following the tables in which the data obtained are listed. 

 For a consideration of the results to determine what conclusions may be 

 drawn from the experimental work, it is necessary to study the data for 

 both seasons as a unit, and in so far as possible to apply statistical methods 

 in this study. 



Production oj dry matter and nitrogen 



A glance at table i (page 134) shows that the different methods of 

 treatment did not result in any evident differences in production for 

 the first period of growth. A comparison of the yields listed in this 

 table with those given in tables 2 and 3, indicates that it is not desirable 

 to turn under the crop at this stage of growth. 



That the effect of lime is evident during the second period of growth 

 is indicated in table 2. In order to decide how conclusive the results 

 were, the probable errors of the means were computed according to 

 Peter's formula,^ 



En, ==±0.8453 



nV 



n-i 



in which T ( + d) denotes the siim of the deviations of every observation 

 from the mean, their signs being disregarded, and n represents the nimiber 

 of observations. In comparing two means the probable error of the 

 gain due to a given treatm.ent was computed according to the formula 



in which Ei is the probable error of one mean and E2 of the other. The 

 results of these computations are given in table 8, in which the different 

 treatments are lettered A, B, and C: 



TABLE 8. Statistical Study of Table 2 



Treatment 



Mean grams of 

 dry matter 



Mean grams of 

 nitrogen 



(A) Unlimed 



(B) Slaked lime 



(C) Ground limestone 



Gain, B over A 



Gain, C over A 



9.21 ± .34 

 II .30 ± .68 

 II .28 ± .14 



0,3216 ± .0060 

 0.4125 ± .0185 

 0.4313 ± .0022 



0194 

 0064 



' A probable error based on two determinations does not necessarily represent strictly the actual 

 frequency diagram, and this must be borne in mind in considering many of the probable errors computed 

 in this bulletin. A discussion of the probable error when the number of observations is small is given 

 in an article in Biomelrika, volume 6, entitled The Probable Error of a Mean. 



