0
$\begingroup$

I have a dataset with one of the variables with exponential distribution. Other variables, including the dependent variable, can be normalized by taking a log.

How can I normalize that one variable?

$\endgroup$
  • $\begingroup$ Why are you transforming your variables? $\endgroup$ – Glen_b Jul 29 '17 at 8:55
  • $\begingroup$ I would like to use the variable for multiple regression. $\endgroup$ – user2728024 Jul 30 '17 at 1:41
  • $\begingroup$ Normality has nothing to do with independent variables, and the (marginal) distribution of the dependent variable is irrelevant; the assumption relates to the conditional distribution of the DV, which you generally can't assess by looking at the marginal distribution (i.e. you cannot tell whether the assumption is reasonable by looking at a histogram or QQ plot of the dependent variable by itself). There are sometimes reasons to transform variables, but this isn't one of them. I expect you've been misled (a number of books get this wrong) $\endgroup$ – Glen_b Jul 30 '17 at 1:46
  • $\begingroup$ Thanks Glen. So for multiple regression, only the dependent variable has to be normally distributed and the independent variable's distribution is irrelevant. Correct me if my understanding is wrong? $\endgroup$ – user2728024 Jul 30 '17 at 6:19
  • $\begingroup$ Right for IV, wrong for DV. Please be very careful of the distinction I made for the DV: the (marginal) distribution of the dependent variable is irrelevant; the assumption relates to the conditional distribution of the DV, which you generally can't assess by looking at the marginal distribution (i.e. you cannot tell whether the assumption is reasonable by looking at a histogram or QQ plot of the dependent variable by itself). .... this is discussed many times on site. You might fruitfully look at (rather than test) the residuals $\endgroup$ – Glen_b Jul 30 '17 at 9:46

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.