standardizing dichotomous variables is particularly silly, since the stan
dev of a dichotomous variable is only a function of the fraction of 1s
(the mean).
the only other idea would be to try to use the person's computer program.
Gary
On Fri, 28 Mar 2003, John Bright wrote:
Yeah, this is a replication. We've tried both
standardization approaches
(standardize raw data then regress and regress then multiply by ratio of
standard deviations) and get the same results but these do not agree with
what the paper says (or what SATA gives us). We do get the same raw
coefficients. So, the problem is in the standardization step and probably
(hopefully) not a silly calculation/coding error.
An additional peculiarity is that most of the independent variables (and all
of the coefficients associated with the variables of interest are
dichotomous. I don't know how one would interpret standardize coefficients
for dichotomous variables.
Any suggestions?
Thanks, john.
On 3/28/03 5:10 PM, "Gary King" <king(a)harvard.edu> wrote:
unless you're replicating someone else's article, i.e., silliness, you
shouldn't be doing this. (tho you can do it by subtracting the mean and
dividing by the sd of each variable). see my how not to lie with
statistics article, at my web page.
Gary
On Fri, 28 Mar 2003, brightjo wrote:
Is there an easy way to coax R into spitting out
standardized coefficients
from an regular linear regression model (from lm)?
Thanks, j.
_______________________________________________
gov2001-l mailing list
gov2001-l(a)fas.harvard.edu
http://www.fas.harvard.edu/mailman/listinfo/gov2001-l