Chi square posthoc?

[Helane Wahbeh]

Hello,

I have have conducted a prevalence study of 25 experiences in three groups: general population, scientists, listserv members.
Many of the experiences have significant chi square values. I want to see if there are significant differences between the three groups (ie. general population vs scientists, general population vs listserv members, scientists vs listserv members).

I've read through the other messages about this but my example is different from those posted in that the categorical variable GROUP is not ordinal.

. tab InfoDream Group, column chi2

+-------------------+
| Key |
|-------------------|
| frequency |
| column percentage |
+-------------------+

| Group
InfoDream | 1 2 3 | Total
-----------+---------------------------------+----------
0 | 161 70 114 | 345
| 56.89 40.23 26.09 | 38.59
-----------+---------------------------------+----------
1 | 122 104 323 | 549
| 43.11 59.77 73.91 | 61.41
-----------+---------------------------------+----------
Total | 283 174 437 | 894
| 100.00 100.00 100.00 | 100.00

Pearson chi2(2) = 69.0182 Pr = 0.000

Any help would be greatly appreciated.

[Joseph Coveney]

Here's one approach.

version 15.0

clear *

input byte InfoDream int (count1 count2 count3)
0 161  70 114
1 122 104 323
end
quietly reshape long count, i(InfoDream) j(Group)

tabulate InfoDream Group [fweight=count]

logit InfoDream i.Group [fweight=count], nolog
pwcompare Group, mcompare(bonferroni) pveffects

exit

[Helane Wahbeh]

Thank you!