1
$\begingroup$

Possible Duplicate:
How should I transform non-negative data including zeros?

I have a continuous variable that has many zeros values and is NOT normal, so I can't use parametric statistics on it.

I tried using ln and root-squared transformations, but they didn't work.

Improve this question
$\endgroup$
4
  • 1
    $\begingroup$ There are many other parametrisations, such as the gamma, beta, or extreme value distributions. Maybe they fit better? $\endgroup$
    – gerrit
    Commented Nov 12, 2012 at 13:30
  • 6
    $\begingroup$ This is NOT enough information to help you. Usually you can't work around zero inflation by just transforming the data. You actually need to model it or use some other sort of appropiate analysis. Telling more about the data (where is it from, how is it distributed) and about your questions might lead to a more helpful answer. $\endgroup$
    – Erik
    Commented Nov 12, 2012 at 13:31
  • $\begingroup$ @Erik is right, we need more information about your situation, your data, your goals, etc. W/o that, we can't help you. Would you mind editing your question to provide this info? You may find this blog post: How to ask a statistics question, & our FAQ helpful in formulating your Q effectively. $\endgroup$
    – gung - Reinstate Monica
    Commented Nov 12, 2012 at 13:51
  • 1
    $\begingroup$ None. This is one of those terrible statistical practices that need to go away. $\endgroup$
    – tchakravarty
    Commented Nov 12, 2012 at 13:54

0

Reset to default

Browse other questions tagged or ask your own question.