Comparing three proportions or medians

[annamaija]

This will be a randomized controlled clinical trial with three groups (1: placebo control, 2: treatment of interest, 3: standard treatment) and the final outcome will be comparison of proportions or medians.

 

Actually the most interesting comparisons would be

1) if there is a difference in proportions (or median scores) btw 1 and 2 and

2) if there is a difference in proportions (or median scores) btw 2 and 3

 

What calculator should I use? Thanks! Anna-Maija   

[Zekeriya Akturk]

Hi Anna. I don't understand how you plan comparing proportions OR medians. What is your outcome variable?
Best wishes.

[annamaija]

The outcome can be 1) proporion of patients (dogs) that improved to a specified level (in certain type of chronic ear infection) or 2) the medians of the summed scores of four measured aspects of the infection (maximum score 12). Maybe even both. And just to mention that all groups (even the placebo) would receive similar mechanical manipulation so there shouldn't be any ethical issues. But of course this would go through ethical committee :-) Anna

[Zekeriya Akturk]

In this case, I would say the numerical data (comparing scores between groups) is a stronger comparison. For this you can use Kruskal Wallis test with post hoc analysis. There are many softwares making this calculations. I am using the SPSS (www.ibm.com/software/analytics/spss/). If you should need help in doing the analysis, send me your data and I will try to do it for you.

[annamaija]

Sorry, I did not ask my question very clearly. I only need help with the sample size calculation. What kind of calculator to use to estimate the sample size for a study with these three groups comparing proportions or means. Anna

    

[Zekeriya Akturk]

http://homepage.stat.uiowa.edu/~rlenth/Power/ of course :)

[annamaija]

Well, my intended original question was that which one of those numerous Rush Lenth calculatores, if any, could I use in this situation.  Anna