If they are a dichotomy (receding vs. not receding and green men vs. blue
men) then the percentages are part of a two by two table and you can make
such a table from the percentages (if you know the sample sizes). Compute
chi-square. Chi-square over N is phi and this is identical to Pearson's
correlation if you make the categories a pair of dummy variables.
alan
Alan C. Acock
[EMAIL PROTECTED]
[EMAIL PROTECTED]
http://www.orst.edu/dept/hdfs/acock/
----- Original Message -----
From: "Ronny Richardson" <[EMAIL PROTECTED]>
To: <[EMAIL PROTECTED]>
Sent: Monday, May 27, 2002 8:26 AM
Subject: Can I Infer Correlation From Proportion?


> I was reading a medical report that gave some values as percentages. It
> said something like 30 percent of all green men have receding hairlines
> while only 5 percent of little blue men have receding hairlines. Is there
> any way I infer anything like a correlation value from that? If not, how
> would I approach significance testing?
>
> I'm thinking I might have to go to nonparametric statistics since none of
> the values being measures (green membership, blue membership, receding
> hairline) are interval scale. Rather, they are all yes/no nominal scale.
>
>
> Ronny Richardson
>
> .
> .
> =================================================================
> Instructions for joining and leaving this list, remarks about the
> problem of INAPPROPRIATE MESSAGES, and archives are available at:
> .                  http://jse.stat.ncsu.edu/                    .
> =================================================================

.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to