This question here Standardized VS centered variables says you can only standardize or center variables after transforming them into log. Because my variables are in different units, my coefficients vary widely from > 10,000 to less than 0. I need to use the scale function on R, but because some of my data is negative, it's impossible to take log of some of the variables. Can I standardize the values without log transformation first or is there any other way to standardize the units?

  • 1
    $\begingroup$ It doesn't say that at all. Rather, if it makes sense you may transform variables before standardizing. You're talking about coefficients in your question, which have nothing to do with standardizing the data. $\endgroup$
    – Nick Cox
    Feb 1 '19 at 13:30
  • 2
    $\begingroup$ Being negative isn't a barrier to transformation, as cube root, asinh and sign(x) log(1 + |x|) are among several possibilities. $\endgroup$
    – Nick Cox
    Feb 1 '19 at 13:52
  • 1
    $\begingroup$ Also, note that you can also transform only part of the variables, as a common solution to the German Credit problem (kaggle.com/uciml/german-credit) suggests. a/w transformation isn't necessary for standartization. $\endgroup$
    – Spätzle
    Feb 1 '19 at 14:28