Two and More Factors ANOVA - Agricultural Statistical Data Analysis Using Stata - page 105

The differences between the varieties occur only with the first
sowing date with a significant difference between the varieties (p =
0.0015), whereas there isn't any difference between the varieties on
the second and third sowing dates.
Another way to view these data is with the contrast command.
This command allows you to view any linear hypothesis involving
factor variables and their interactions. Enter the following command
immediately after the factorial ANOVA:
contrast variety @ date
which results in the following output:
Contrasts of marginal linear predictions
Margins : asbalanced
------------------------------------------------
| df F P>F
-------------+----------------------------------
variety@date |
1 | 3 19.72 0.0000
2 | 3 0.91 0.4457
3 | 3 0.00 0.9998
Joint | 9 6.88 0.0000
|
Residual | 33
------------------------------------------------
This command calculates the probabilities of seedstem differences
between the varieties for each of the sowing dates. Notice the F values
are different from the ANOVA tables calculated for each date above.
This is because these F values are calculated using the mean square
for the residuals (25.6691919) from the overall ANOVA table as the
denominator rather than the mean square for the residuals from the
individual ANOVA tables by date. The overall mean square for the
residuals is a more appropriate denominator if the individual residual
mean squares are all similar. This overall residual mean square is based
on larger degrees of freedom, therefore, it will be a smaller value than
for the individual ANOVAs, which means it has greater power to
detect differences. If, however, the individual ANOVAs have residual
mean squares that are quite different from each other, as in this case,
then the individual ANOVAs would be more appropriate.