1
$\begingroup$

I've alway wondered if you are given tabular data where things have different distributions (some features are normally distributed, some are log-normally distributed) if centering the data by...

$$ \frac{x - \mu}{\sigma} $$

Do we destroy the information about the variables which do not have a normal distribution?

Improve this question
$\endgroup$

1 Answer 1

Reset to default
1
$\begingroup$

If your level of measurement is an interval scale, you can do positive linear transformations ($x' = ax+b$ with $a>0$) without losing information. This does not depend on the distribution of your data.

Note that for centering, you can set $a = \frac{1}{\sigma}$ and $b = \frac{\mu}{\sigma}$. Since $\sigma>0$, this will be a positive linear transformation.

Improve this answer
$\endgroup$
2
  • $\begingroup$ Thanks for your answer, what qualifies as an "interval scale?" $\endgroup$
    – Joff
    Apr 27, 2020 at 12:04
  • $\begingroup$ When you can interpret the difference between values in a meaningful way. If you can compute mean and standard deviation, you should typically have an interval variable. You can find more information on levels of measurement for instance here $\endgroup$
    – mesolimbic
    Apr 27, 2020 at 12:56

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.