Post Hoc Tests - Agricultural Statistical Data Analysis Using Stata - page 152

rounded to the nearest whole number, 2 in this case. This means there
is the chance of finding two significantly different comparisons when,
in fact, there are none. All possible comparisons together are often
referred to as a family of comparisons and the Type I error rate in this
case as the familywise error rate.
There are several methods available to evaluate all the pairwise
comparisons. Load the dataset Onion Small Trial 1999.dta, which
is a small onion variety trial. Then analyze the data with the anova
command. Follow this with the pwcompare command. he com-
mands and output are shown below:
anova yieldacre entry rep
pwcompare entry , pveffects
Number of obs = 15 R-squared = 0.9178
Root MSE = 78.4047 Adj R-squared = 0.8561
Source | Partial SS df MS F Prob > F
--------+----------------------------------------------------
Model | 548744.502 6 91457.4169 14.88 0.0006
|
entry | 532527.14 4 133131.785 21.66 0.0002
rep | 16217.3618 2 8108.68091 1.32 0.3198
|
Residual| 49178.37 8 6147.29625
---------+----------------------------------------------------
Total | 597922.872 14 42708.7765
Pairwise comparisons of marginal linear predictions
Margins : asbalanced
-----------------------------------------------------
| Unadjusted
| Contrast Std. Err. t P>|t|
-------------+---------------------------------------
entry |
2 vs 1 | 78.65 64.01717 1.23 0.254
3 vs 1 | -10.40599 64.01717 -0.16 0.875
4 vs 1 | -372.438 64.01717 -5.82 0.000
5 vs 1 | -339.768 64.01717 -5.31 0.001
3 vs 2 | -89.05599 64.01717 -1.39 0.202
4 vs 2 | -451.088 64.01717 -7.05 0.000
5 vs 2 | -418.418 64.01717 -6.54 0.000
4 vs 3 | -362.032 64.01717 -5.66 0.000
5 vs 3 | -329.362 64.01717 -5.14 0.001
5 vs 4 | 32.67 64.01717 0.51 0.624
-----------------------------------------------------