2
$\begingroup$

I am looking for a (commonly used) probability density function, which would look like a normal distribution flipped upside down. It would look like a uniform distribution with a dent in the middle.

Just to be clear, I am dealing with a continuous random variable within some range, say [-1,1].

Sometimes I have data which is concentrated around zero, but other times I have data which is concentrated toward 1 and -1, while the region around zero is relatively underpopulated. Is there a kind of pdf which (depending on some parameter) can represent these two cases?

Improve this question
$\endgroup$

1 Answer 1

Reset to default
4
$\begingroup$

Beta distribution? You'd have to rescale your [-1,1] into [0,1].

enter image description here

Don't know if this will get the level of weight you want into the corners. Decreasing the shape parameters forces more weight to them though.

Improve this answer
$\endgroup$
3
  • $\begingroup$ Actually, the shape parameters are the exact parameters I need to adjust to fit both cases I mentioned. It looks like this is exactly what I was looking for! $\endgroup$
    – Angelorf
    Commented Feb 10, 2014 at 10:32
  • 6
    $\begingroup$ The beta distribution illustrated above has an antimode at $x = 0.5$. This kind of minor terminology polarises people all the way from "yet another term we don't really need" to "just the term I was looking for". $\endgroup$
    – Nick Cox
    Commented Feb 10, 2014 at 11:36
  • 1
    $\begingroup$ Well, the term is certainly better than 'abyss'! $\endgroup$
    – Angelorf
    Commented Feb 10, 2014 at 17:20

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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