[R] simple on point bisserial correlation

Dmitry Gospodaryov gospodaryov at rambler.ru
Wed May 5 21:54:11 CEST 2010


Dear developers of R-project,
I have such data
height of male persons
181, 178, 182, 160, 187, 193, 184, 184,
175, 178, 184, 184, 174, 185, 175
height of female persons
164, 165, 160, 174, 167, 161, 164, 165,
169, 175, 165, 155, 172, 164, 172, 166,
160, 159, 158, 173
I want to determine if height is dependent on
gender, and I need to find, as I understand, point
biserial correlation. I had tried to use package "ltm",
and obtained coefficient which is about 0.74. I also
had tried to calculate it manually (with some MS Excel help
for simple calculations), using equation given in Internet
(e.g., in Wikipedia; this equation is also given in some
copy-books) and obtained coefficient is about 0.77.
"ltm" provides somewhat other equation with pi-numbers.
So, the questions are following:
1). Do exist packages in R-project, except of "polycor" and
"ltm", which provide calculation of point biserial correlation?
1a). Why "ltm" use "non-traditional" equation for
bisserial coefficient? Which is significance of this difference?
2). Can I also calculate significance of the biserial correlation
coefficient (p < ???; using, probably, t-test) by means of
R-project?

Thank You for advance.
With regard,
Dmitry Gospodaryov.



More information about the R-help mailing list