pairwise comparison of proportions

[Lissa]

If someone could help me with this question, I would extremely appreciate the help since I've been trying to find the answer for days by now...

I want to run multiple chi-square tests (or logistic regressions) and adjust for multiple comparisons. so basically, i want to run pairwise comparison of proportions.

I would want to run either chi-square analyses with Bonferroni adjustment or logistic models (GLM) with scheme adjustment... However, when I search for the syntax, i only find codes for pairewise comparison for linear models (i.e., ANOVAs with adjustment)... Would anyone know the code for categorical variables? For example, how do I run chi-square analyses with bonferroni adjustment in STATA?

Thanks so much in advance!

 

[bukharin]

If you're running Stata 12 you can use -pwcompare- after a logistic model

 

[Lissa]

thats so much for your reply. I am running state 11 though... do you know if there is a similar command for state 11? if not, i'll try downloading stata12 - wasn't successful with it till now...

also, if i am running multiple logistic models whereas i am systematically changing my reference category (which leads to the multiple comparison problem), would the pwcompare command adjust for that?

thanks so much again, i really appreciate!

 

[bukharin]

No, -pwcompare- was introduced in Stata 12. As far as I know you can't download Stata - you need to buy it either from StataCorp or from a distributor.

With -pwcompare- you don't need to run multiple models with different reference categories - you just run one model, then run -pwcompare- and it automatically compares all categories and optionally adjusts the P-values / CIs for the multiple comparisons.

 

[Lissa]

oh, that is amazing. last question, if that is ok - what is the statistical test that pwcompare does to adjust the p-values?

thanks again.

 

[bukharin]

It offers a few different methods. You can read the help file here

 

[Lissa]

thank you very much. i will certainly get state 12. thanks

 

[bukharin]

Lissa, this comment is not directed at you but is instead a general comment to clear up any confusion for people reading the thread:

As far as I know you can't download Stata

Not officially anyway; I wouldn't be surprised if there were non-official versions floating around. However, using a non-official version of Stata is a bad idea