[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: Cramers V with weighted data

From   Steven Samuels <>
Subject   Re: st: Cramers V with weighted data
Date   Thu, 14 Aug 2008 12:55:50 -0400

I have to apologize to everyone. The statistic computed by the formula that I gave is Cramer's Phi statistic, which is Cramer's V for 2x2 tables. Thanks to Nick Cox for pointing out my error and pointing me to the proper definition.

The proper formula for Cramer's V is:

di "Cramer's V = " sqrt(e(cun_Pear)/(e(N)*min((e(r)-1),(e(c)-1))))

On Aug 14, 2008, at 10:47 AM, Waldo, Daniel R. (CMS/ORDI) wrote:

In the case of complex survey design, should the denominator of the
calculation take the design effect into consideration? That is, should I
use N/DEff as opposed to N?

Date: Tue, 12 Aug 2008 08:34:30 -0400
From: Steven Samuels <>
Subject: Re: st: Cramers V with weighted data

I did.

Thanks, Paul.

- -Steve
On Aug 12, 2008, at 5:21 AM, Paul Seed wrote:

I think Steven meant:

	gen crV= sqrt(e(cun_Pear)/e(N))


	di "Cramer's V = " sqrt(e(cun_Pear)/e(N))

Paul Seed, Lecturer in Medical Statistics
KCL School of Medicine, Division of Reproduction and Endocrinology
tel  (+44) (0) 20 7188 3642

On Aug 8, 2008, at 12:21 PM, Steven Samuels wrote:
-svyset- the data with pweights equal to your weight. Run -svy:
tabulate- and divide the uncorrected Pearson Chi Square statistic
by the sample size.  You can program this by dividing the returned
scalar e(cun_Pear) by the scalar e(N)

gen crV= e(cun_Pear)/e(N)


*   For searches and help try:
*   For searches and help try:

© Copyright 1996–2015 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index