"Paul C. Smith" wrote:
> John W. Kulig wrote:
> > Convert each of the chi-squares to an effect
> > size measure such as the contingency coefficient. i.e. C = square root
> (chi
> > square/(N + chi square)). C will range from 0 to (close to) 1. Then you
> can
> > rank order the 32 behaviors in terms of the C - and see if the rank
> ordering
> > makes sense.
>
> Intriguing. Do you have a source (e.g., textbook) describing this?
>
Siegel, S. (1956) _Nonparametric Statistics for the Social Sciences_
McGraw-Hill. pages 196-201.
Lehman, R.S. (1995) _Statistics for the Behavioral Sciences_Brooks/Cole.
pages 309-310.
I bet most stat books with decent nonparametric sections will have it as
well.
Lehman also discusses the phi coefficient = (square
root(chi-square/(N(k-1)))) where k = rows or columns whichever are smaller.
It, like C, is a "correlation-like" measure of association for categorical
data. I could be wrong about this, but I believe if you simply run categorical
data (e.g. dummy codes 0,1) into a Pearson r, you "get" the phi coefficient,
and pretty much the same value you'd get by doing a chi-square and then
converting using this formula. Phi and C measure the same thing, but act
differently, I believe, as you get closer to a correlation of 1. Whatever
Kristen uses, however, should be determined primarily by her research
questions and not pure statistical matters. I assume she has a 2x2
(student/professor, and "it's cheating/it's not cheating" jusgments) and that
she is interested in differences between the 32 behaviors. I'm a big booster
of looking at data descriptively before hypothesis testing is done; that's why
I brushed aside the Type I error problem (I really do worry about Type I
errors, but I think they should be balanced against the liklikehood of other
types of errors).
btw, one of the reasons you have to convert chi-squares into contingency
coefficients is that the size of the obtained chi-square is dependent on the
size of the table and N. So you cannot conclude that a chi square of 30
necessarily means a stronger relationship between the 2 variables than a chi
square of 20. Her tables and N may be the same for all 32 behaviors, but it is
still nice to convert to a effect size or contigency measure if one wants to
get a handle on the degree of association between the 2 variables.
--
---------------------------------------------------------------
John W. Kulig [EMAIL PROTECTED]
Department of Psychology http://oz.plymouth.edu/~kulig
Plymouth State College tel: (603) 535-2468
Plymouth NH USA 03264 fax: (603) 535-2412
---------------------------------------------------------------
"The only rational way of educating is to be an example - if
one can't help it, a warning example." A. Einstein, 1934.
