0

I have the histogram in the picture:

enter image description here

I would like to compute the mode. I fit first the distribution with a beta distribution and then I compute the mode. Is there a better approach?

I would appreciate if you could add MATLAB code.

Nick Cox
  • 48,377
  • 8
  • 110
  • 156
gabboshow
  • 641
  • 6
  • 17
  • 2
    Possible duplicate of [How to find the mode of a probability density function?](http://stats.stackexchange.com/questions/176112/how-to-find-the-mode-of-a-probability-density-function) – Nick Cox Jun 06 '16 at 07:08
  • 5
    You've identified a fitted mode from a fitted beta distribution. The $x$ axis tick marks of your graph are obscured by the histogram bars (poor choice by or in MATLAB) but by eye from your graph the fitted mode is about 0.7 whereas values more like 0.8 seem more plausible. So other more data-driven methods of estimating the mode seem attractive. See other comment for one link. Incidentally, the secondary mode at about 0.25 is too pronounced to be easily dismissed as sampling fluctuation and is thus inconsistent with a model of a single beta distribution. – Nick Cox Jun 06 '16 at 07:14
  • 2
    I think that the mixture of data and fitted curve distinguish this question somewhat from the proposed duplicate - part of the issue here is the appropriateness of fitting a beta, whereas the issue of finding the mode of the fitted beta is essentially trivial. (The proposed duplicate is more about finding the mode for a "trickier" theoretical distribution.) – Silverfish Jun 06 '16 at 08:20
  • @Silverfish It's not clear to me what else can be said about fitting a beta distribution here. The fit could be a lot worse and it could be a lot better. The question cited includes answers about getting modes from any dataset. – Nick Cox Jun 06 '16 at 16:30

0 Answers0