On Sat, Jan 29, 2011 at 1:23 PM, Joshua Wiley <jwiley.ps...@gmail.com> wrote: >> xxAB <- dat$xxA * dat$xxB >> meanAB <- meanA * meanB >> sdAB <- sqrt(sdA^2 + sdB^2 + meanA^2 * sdB^2 + meanB^2 * sdA^2)
don't know why I did that the hard way (besides I think I messed up the formula slightly), but the point is there's still a difference between the scaled interaction from unscaled variables and the interaction of scaled variables. sdAB <- sd(xxAB) # <-- what I should have done > (xxAB - meanAB)/sdAB [1] 0.18412267 -0.97105894 -1.45088077 -1.37131900 0.09339332 0.74663645 [7] -1.27092403 -0.41183471 -1.45890639 0.06568750 0.60130106 1.35945072 [13] 0.79255938 1.80314145 0.51553234 -0.23553700 -0.42045914 -0.73543976 [19] 1.40009474 0.68251640 >> (sxxAB <- with(dat, scale(xxA) * scale(xxB))[1:20,]) > [1] 0.01838999 0.25899229 -0.15337178 1.07536361 -0.26520706 0.25119036 > [7] -0.11295999 0.05205466 -1.49676116 -0.14367368 -1.86921225 0.80860167 > [13] -0.44816896 1.44122535 -0.73361472 -0.14930187 -1.46192081 0.19784309 > [19] 0.62089372 0.0821291515 ______________________________________________ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
