Hi all, When I use "lda" for discriminant analysis, should I normalized my data (variables) to mean 0, variance 1 before running "lda" if my variables might not be exactly on the same scale? I have this question because in principle component analysis, this is indeed an issue where we can choose covariance matrix or correlation matrix when running the analysis. Just wondering if this is also an issue people should consider when using discriminant analysis. second question: "coef(lda object)" returns coefficients of the variables, are they standardized coefficients (i.e. they should be applied to standardized variables to calculate discrimnant scores), or are they unstandardized coefficients (i.e. they can be directly applied to the original variable to compute the discrimnant scores)? SPSS can output both coefficients, but it seems that "lda" can only output one, I think they should be standardized coefficeients, but not sure. Thanks very much