Test of difference in median

[Rochelle Zhang]

Dear all,
I want to test difference in median between two groups – those with eventid=1 vs. 0

I googled and find this stata journal article, “Parameters behind "nonparametric" statistics: Kendall's tau, Somers' D and median differences”.


pretend data :

clear
input double Mdays byte eventid
63.04238 1 
72.77879 0 
72.77879 0 
72.77879 1 
72.77879 0 
68.61013 1 
69.21858 1 
end

I tried
cendif Mdays, by(eventid )

but do not know if I am on the right track, and how to read the output.

Thanks,

Rochelle

[Marcos Almeida]

You didn't say much about the variable Mdays. If it is "processed" information concerning, well, days, I gather you should think about count data analysis, Poisson-like models.

With regards to median, you may think, theoretically speaking, about - ranksum - or - median - test.


That said, in case we have observational data, unadjusted tests such as these ones are mainly used as an exploratory basis.

Description

ranksum tests the hypothesis that two independent samples (that is,
unmatched data) are from populations with the same distribution by using
the Wilcoxon rank-sum test, which is also known as the Mann-Whitney
two-sample statistic (Wilcoxon 1945; Mann and Whitney 1947).

median performs a nonparametric K-sample test on the equality of medians.
It tests the null hypothesis that the K samples were drawn from
populations with the same median. For two samples, the chi-squared test
statistic is computed both with and without a continuity correction.

Indeed, if both groups have the same pattern of variance, I believe both tests would provide similar results.

To end, - cendif - "Robust confidence intervals for median and other percentile differences between two groups", is a user-written SJ program, as you are expected to underline. Its author is Roger Newson, an active member in this Forum.

[Rochelle Zhang]

Thanks Marcos. You are quite helpful !

Rochelle

[River Huang]

In Conroy's paper, http://www.stata-journal.com/article...article=st0253, you can run a quantile regression to test the equality of medians between two groups (page 188).

qreg Mdays eventid