187 



If we average these results as they stand, C will be given two plots 

 in the poor row 1, and B two plots in the good row 3. The average of 

 B will thereby tend to be made greater than the average of C. This will 

 not do. Where any variety had more than one plot in a row, the figures 

 for these plots were therefore replaced by their mean, and the average 

 calculated from the four resulting figures. In this way we get the com- 

 parable or standardised yields for B, C, D and ( > : 



The yields of B, C and D are in the same order as size of set, and the 

 difference (7-3 lbs.) between the greatest and least is over five times the 

 probable error of the dift'erence (1-4) — taking this probable error as based 

 in effect on only four pairs of plots. 



But there remain A, E and F of the variety used for testing the effect 

 uf size of set. A has plots only in rows ], 3 and 4: hence the averages 

 obtained in the same way as above for A and for the plots in the same 

 rows of B, C and D are not comparable with the standardised figures 

 above; and we want a comparable figure. 



This was obtained in the following way: on the three pairs of plots 

 available, the total yield of A was 76-5, and of B 111-2. The standardised 

 yield of B is 34-3. An estimated standardised yield of A may therefore 

 be put at 76-5 x 34-3 ^ 111-2 or 23-6. From the similar figures for C, 

 D and G we obtain estimates of 24-5, 24-3 and 24-4. These figures are all 

 fairly close together, and we may take their mean, 24-2, as a fairly close 

 estimate of the standardised yield of A. Standardised estimates for E 

 and F were similarly obtained, viz. E, 47-5. and F, 67-3. 



These figures increase without a break from the smallest sets A, to 

 the largest sets F, so there can be no doubt about the result. For B, C 

 and D, as already stated, the probable error of a difference is about 

 1-4 lbs. For A, E and F. owing to the method of estimation, we cannot 

 state a precise probable error, But if the probable error of the difference 



