Simple answer: Yes.
See the following for an example:
The original reference that provides this test is:
A Ranking Variable Analogue of Biserial Correlation:
Implications for Short-Cut Item Analysis
Gene V. Glass
Journal of Educational Measurement , Vol. 2, No.
1 (Jun., 1965), pp. 91-95
Article Stable URL:
http://www.jstor.org/stable/1433839
N refers to the total number of persons.
However, I wonder why you say "paired" since the rank biserial
correlation
assumes that the groups are independent (if we represent groups by X,
then X is a dichotomous variable representing the underlying ranking
and
is correlated with some continuous variable Y).
You should perhaps be more concerned with whether you have ties and
how to deal with them.
-Mike Palij
New York University
----- Original Message -----
Sent: Tuesday, March 05, 2013 12:11
PM
Subject: FW: Rank Biserial
correlation
To Whom This May Concern,
I am attempting to calculate
the rank-biserial correlation between two sets of data which have been grouped
based upon a dichotomous variable. I have used the equation:
(mean rank group 1 mean rank group 2) *2/ N.
Does this look
correct to you and does N refer to the total sample size or the number of
paired observations (I have two groups of 42, thus 84 participants in
total)?
Your help with this
matter would be extremely appreciated.
Kind
regards,
Mr Edd
Thomson
Locked
