362 EXPERIMENTS WITH FERTILISERS [chap. 



Table Q\\\.— Continued. 



yielded less than Plot 3, while in a single year Plot 12 

 has risen as high as 196 per cent, or fallen as low as 90 

 per cent, of Plot 3. 



Applying what is known as the method of least 

 squares to the results, we can calculate that the mean 

 error of a single result is ± 10 per cent, and that the 

 probable error of the fifty years' mean is ± 1-9 

 per cent. In other words, Plot 12 is probably better 

 than Plot 3 by more than 8-i, and less than 11-9 per 

 cent ; but this superiority could never be assured from a 

 single year's experiment, because it is smaller than the 

 mean error, which is equal to a 10 per cent difYerence 

 between the two plots. The probable error is always 

 reduced by the number of trials ; if the fifty years are 



