0

Given the histogram below, what common distribution would be well fit to this data i.e. Beta, skewed normal. I want to use the data as a prior in Bayesian analysis so want to approximate it by a distribution in the exponential family to make this easier to do analytically.

Histogram

We can ignore the high values at 0 and the value at 0.75 as they are errors/outliers.

rwolst
  • 741
  • 1
    The histogram itself. Your question doesn't really make sense unless you have some model in mind about how the numbers were generated. – Raskolnikov Jul 15 '13 at 10:25
  • I was wondering whether it looked like any commonly studied distributions. The data is necessarily between 0 and 1. I was initially thinking of a Beta distribution but when I estimated the parameters with my data it was not such a good fit. – rwolst Jul 15 '13 at 10:42
  • 1
    One problem with such questions is that if you choose a different bin size / bandwidth, you might get a very different plot out. In particular, the little mode at 0.1 is of interest ... and may suggest you have a bimodal distribution ... which may become more apparent with a smaller bandwidth ... or appear to disappear with more smoothing with a larger bandwidth. – wolfies Jul 15 '13 at 13:55

0 Answers0