It can be seen immediately that the means in each (block) column are not zero. This requires that we sweep
again to eliminate differences between blocks. Why has this happened? If we look back at the block means
first in the complete case and then in the incomplete case we see that the means for block 1 are
552(complete) and 520(incomplete). The mean for the incomplete case is too low because treatment D,
which is the best treatment, was missing in block 1. Hence, we have not fully allowed for the high level of
yields in block 1 and the further sweeping will do this.

So we sweep again by blocks.


Mean +31 +24 -10 -44

 -8 -11 -27
 -10 -3 +2
 +17 -2 +9
 +14 +28 -11


Now, for exactly the same reason, we find that the means for the treatments are not zero. So we sweep
again by treatment and continue to sweep alternately by block and treatment until we achieve residuals
which give zero means for blocks and treatments.


Treatment
Means
-15 +7 +4 -12
-4 -6 +1 +6
+8 +9 -10 +1
+10 +4 +18 -21


Means


Means