-1
$\begingroup$

enter image description here

What type of distribution does this most closely resemble? I was thinking that this graph didn't really resemble any type of distribution.

$\endgroup$
4
  • $\begingroup$ What kind of data are these? Discrete or continuous? What is the context? $\endgroup$ Jul 27 at 5:32
  • $\begingroup$ Looks like a mixture, reminds me of a tweedie distribution: en.wikipedia.org/wiki/Tweedie_distribution $\endgroup$
    – Tylerr
    Jul 27 at 19:48
  • $\begingroup$ Given there's an infinite number of distributions that are arbitrarily close to the data, without some restrictions, there's no useful way to answer the question beyond the one that matches the data exactly (i.e. the ecdf itself), which comes as close to the distribution of the data as you can get. Beware also of picking a distribution out of some arbitrary laundry list (as many programs have been written to do). Much better to think about (a) whether you need to choose one at all; (b) what you know about the variable (such as its support), ... $\endgroup$
    – Glen_b
    Jul 27 at 23:08
  • $\begingroup$ (c) whether the data you have represents a sample that it makes sense to apply a distribution to; (d) what models may have been used by experts for similar data previously, and (e) consider what you're trying to achieve by doing so (since that may well guide choices of distribution) $\endgroup$
    – Glen_b
    Jul 27 at 23:08
2
$\begingroup$

This appears to come from https://rdrr.io/rforge/Lahman/man/Pitching.html, so sports data. One row per player per season since 1950. GS is games started.

This distribution probably doesn't have any nice name, because the distribution itself is not nice. That doesn't mean that it doesn't follow any distribution. Just describe it in some useful qualitative terms: it's bimodal, with approx 25% of the sample at 0, and a small local mode at 33. The max is (approx) 36. Really, the histogram (more of a barplot actually) tells the whole story.

$\endgroup$
1
  • 2
    $\begingroup$ It could perhaps be described as a mixture of two Poissons or Negative Binomials, possibly a three-component mixture with a point mass at 0. $\endgroup$ Jul 27 at 6:21

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.