Question on oneway anova and pairwise comparison

[Pei ZHI]

Recently, I'm conducting research, we would like to see some difference between different national identity groups. First, I applied One-way ANOVA (oneway) command to test the mean comparison among identity groups. But when I test it with pwmean command, the results changed. Does anyone know the reason and the difference between those two commands?

oneway polsati newnid5, bonferroni tabulate

new coded |
national | Summary of political satisfactory
identity | Mean Std. Dev. Freq.
------------+------------------------------------
1 | 4 1 198
2 | 4 1 115
3 | 4 1 448
4 | 4 1 50
5 | 4 1 124
------------+------------------------------------
Total | 4 1 935

Analysis of Variance
Source SS df MS F Prob > F
------------------------------------------------------------------------
Between groups 29.9813114 4 7.49532785 8.25 0.0000
Within groups 844.664678 930 .908241589
------------------------------------------------------------------------
Total 874.645989 934 .936451809

Bartlett's test for equal variances: chi2(4) = 33.8585 Prob>chi2 = 0.000

Comparison of political satisfactory by new coded national identity
(Bonferroni)
Row Mean-|
Col Mean | 1 2 3 4
---------+--------------------------------------------
2 | -0
| 1.000
|
3 | -0 -0
| 0.000 0.219
|
4 | -0 -0 -0
| 0.010 0.202 1.000
|
5 | -1 -0 -0 -0
| 0.000 0.015 0.908 1.000



Then, I tried pwmean to see the mean comparison, but I got the different answer like below:

. pwmean polsati, over(newnid5) mcompare(bonferroni) cieffects pveffects effects cimeans

Pairwise comparisons of means with equal variances

over : newnid5

--------------------------------------------------------------
| Unadjusted
polsati | Mean Std. Err. [95% Conf. Interval]
-------------+------------------------------------------------
newnid5 |
1 | 4.419192 .067728 4.286275 4.552109
2 | 4.295652 .0888693 4.121245 4.47006
3 | 4.066964 .0450258 3.9786 4.155328
4 | 3.92 .134777 3.655498 4.184502
5 | 3.903226 .0855835 3.735267 4.071185
--------------------------------------------------------------

---------------------------
| Number of
| Comparisons
-------------+-------------
newnid5 | 10
---------------------------

-----------------------------------------------------
| Bonferroni
polsati | Contrast Std. Err. t P>|t|
-------------+---------------------------------------
newnid5 |
2 vs 1 | -.1235397 .1117355 -1.11 1.000
3 vs 1 | -.3522276 .081329 -4.33 0.000
4 vs 1 | -.4991919 .1508374 -3.31 0.010
5 vs 1 | -.5159661 .1091403 -4.73 0.000
3 vs 2 | -.2286879 .0996247 -2.30 0.219
4 vs 2 | -.3756522 .1614391 -2.33 0.202
5 vs 2 | -.3924264 .1233786 -3.18 0.015
4 vs 3 | -.1469643 .1420991 -1.03 1.000
5 vs 3 | -.1637385 .096705 -1.69 0.908
5 vs 4 | -.0167742 .1596539 -0.11 1.000
-----------------------------------------------------

even the mean of each identity groups are different in these two commands. Does anyone know how to deal with it? Thanks!

[Carlo Lazzaro]

Pei:
welcometo this forum.
If your regressand is continuous, there's nothing that -regress- can't do better than -anova-.
Then you can use -contrast- to delve into multiple comparison issues.
As an aside, please read and act on the FAQ to post more effectively and increase your chances of getting helpful replies. Thanks.