Partition of Sums of Squares in ANOVA (II)

Each of the blocks above contain sums of squares that summed over all values of i and j represent different sources of variation: total, between and within treatment variation.

In ANOVA we partition the total sum of squares (TSS) in:

Sums of squares due to treatments (SSTreat):

measures variation due to differences between treatments (explained variation).

with degrees of freedom t-1

Sums of squares error (SSE):

measures random variation within treatments (unexplained variation).

with degrees of freedom N-t