Bonferroni after pairwise comparisons after "meologit"

[Gianfranco Di Gennaro]

Hi everybody. I apologize if I bother you.

I'm doing a study about which I've already gotten help from Joseph Coveney and Clyde Schechter (I thank you again).

I have an ordinal outcome (symptom severity that increases from 0 to 5) and which is evaluated in 4 different timepoints (30, 60, 120, 180 minutes) 4 drug doses.

meologit v i.dose##i.time || id:, or

I am interested in making pairwise comparisons between all crossed doses at each of the 4 timepoints in order to evaluate the dose-response.

So I'm interested in seeing the overall odds/ratio of having higher severity in one dose than the other. For example, in dose 2 versus 1, and so on.

I used the following command:
pwcompare i.dose##i.time, pveffects or mcompare(bonferroni) cim eff

Is it correct? I also wonder: how does Bonferroni's correction take place in this case? Are the unadjusted p multiplied by ALL listed comparisons?

I'm interested in multiplying by 10 (I have ten pairwise comparisons for each timepoint).

To do this the only option is to have unadjusted p-values with
pwcompare i.dose#i.time, pveffects or ci eff
and multiply them by 10 manually?

Thank you.
I'm afraid I duplicated the content because I've already asked about this study elsewhere, but it was difficult to relate to other posts to relate to this request.

My data are below

* Example generated by -dataex-. For more info, type help dataex
clear
input byte dose int time byte(rat v) float id
0  30 1 4  1
0  60 1 5  1
0 120 1 5  1
0 180 1 5  1
0  30 2 5  2
0  60 2 5  2
0 120 2 5  2
0 180 2 5  2
0  30 3 5  3
0  60 3 5  3
0 120 3 5  3
0 180 3 5  3
0  30 4 4  4
0  60 4 4  4
0 120 4 4  4
0 180 4 5  4
0  30 5 4  5
0  60 5 5  5
0 120 5 5  5
0 180 5 5  5
0  30 6 3  6
0  60 6 4  6
0 120 6 5  6
0 180 6 4  6
0  30 7 4  7
0  60 7 5  7
0 120 7 5  7
0 180 7 5  7
0  30 8 5  8
0  60 8 5  8
0 120 8 5  8
0 180 8 5  8
1  30 1 4  9
1  60 1 5  9
1 120 1 5  9
1 180 1 5  9
1  30 2 3 10
1  60 2 5 10
1 120 2 5 10
1 180 2 5 10
1  30 3 5 11
1  60 3 4 11
1 120 3 5 11
1 180 3 5 11
1  30 4 4 12
1  60 4 5 12
1 120 4 5 12
1 180 4 4 12
1  30 5 4 13
1  60 5 4 13
1 120 5 4 13
1 180 5 4 13
1  30 6 4 14
1  60 6 5 14
1 120 6 5 14
1 180 6 5 14
1  30 7 4 15
1  60 7 5 15
1 120 7 5 15
1 180 7 5 15
1  30 8 5 16
1  60 8 5 16
1 120 8 4 16
1 180 8 5 16
2  30 1 2 17
2  60 1 2 17
2 120 1 5 17
2 180 1 5 17
2  30 2 2 18
2  60 2 3 18
2 120 2 5 18
2 180 2 5 18
2  30 3 2 19
2  60 3 3 19
2 120 3 5 19
2 180 3 3 19
2  30 4 3 20
2  60 4 4 20
2 120 4 4 20
2 180 4 5 20
2  30 5 2 21
2  60 5 3 21
2 120 5 5 21
2 180 5 5 21
2  30 6 2 22
2  60 6 3 22
2 120 6 5 22
2 180 6 4 22
2  30 7 1 23
2  60 7 3 23
2 120 7 5 23
2 180 7 5 23
2  30 8 2 24
2  60 8 3 24
2 120 8 5 24
2 180 8 5 24
3  30 1 0 25
3  60 1 2 25
3 120 1 2 25
3 180 1 2 25
end

[Carlo Lazzaro]

Gianfranco:
I would consider:

pwcompare i.dose##i.time, pveffects mcompare(bonferroni)

[Gianfranco Di Gennaro]

Thank you Carlo Lazzaro . But i.dose##i.time will produce tens of comparisons. Will p values will be multiplied for all these tens of comparisons?
Thanks again.

[Carlo Lazzaro]

Gianfranco:
looking at the p-values with and without the Bonferroni's correction, the answer seems to be adfirmative.