256 



University of California Publications in Agricultural Sciences [Vol.3 



It will be noted that the probable error of a single nitrate deter- 

 mination is 2.7 ± .09 milligrams or 3.2 per cent of the amount deter- 

 mined. The error for a single determination, expressed on the field 

 samples as a percentage, is 17.4 or about 5.1 times greater than the 

 laboratory error, even after making allowance for this error. With 

 sixteen determinations, the probable error becomes 2.7 ± .022 milli- 

 grams or 0.8 per cent. Further, with eighty-one samples, the probable 

 error is only 2.7 ± .009 milligrams or 0.3 per cent. 



In allowing for this laboratory error in the various series reported, 

 it is evident that it becomes relatively larger the smaller the amount 

 of nitrate determined within the limits of the amounts found in the 

 present study. In other words, the probable error will remain the 

 same while the amount of nitrate determined decreases, so that the 

 probable error will form a larger percentage of the determination. 

 On the other hand, the probable error becomes relatively smaller as 

 we increase the actual amount of nitrate nitrogen again within the 

 limits of the amounts reported here. The same is true for the co- 

 efficient of variability. This increased ratio is brought out in table 8, 

 which is simply a reconstruction of table 5, after making allowance 

 for the laboratory error. 



The coefficient of variability computed from the standard deviation 

 (.13 ± .01 mg.) and the average amount, of nitrate nitrogen reported 

 for the various series is given in column one. This figure is the largest 

 for the subsoil samples in the field and the smallest for the surface 

 samples treated with blood, the mean of the eighty-one readings being 

 0.7 and 21.0 milligrams, respectively, in the two cases. In the second 

 column, the corrected coefficients of variability for the field samples 

 are given, these being the difference between coefficients of variability 

 before the laboratory error was allowed for and the various coefficients 

 of variability given in column one. It will be noted that the same 



