1

Hi I'm trying to estimate the data distribution using Matlab.

For one dimensional data, I can definitely use ksdensity.

However my problems is that I need multi-dimensional joint distribution and conditional distribution.

I've tried kde tools from UCI. It is not functioning in my case and I cannot figure out why. So I'm asking for another tool I can possibly use..

Edit

The toolbox is not working and giving extreme results. I used 1e5 points and it might because the points are too dense.

KDE toolbox result

KDE toolbox

ksdensity result

enter image description here

SolessChong
  • 3,370
  • 8
  • 40
  • 67
  • 1
    This may help: http://stackoverflow.com/questions/18639518/generate-and-plot-the-empirical-joint-pdf-and-cdf-in-matlab – Luis Mendo Dec 26 '13 at 12:36
  • If you don't have any model of the distribution, then multidimensional kernel density estimation is the way to go. Can you explain how exactly the "kde tools" are not working? How many data points do you have and how many dimensions? – A. Donda Dec 26 '13 at 15:18
  • @A.Donda I updated the question and posted a result. – SolessChong Dec 27 '13 at 05:00
  • 1
    Thanks, but from these plots it's still hard to say what is the problem. How does the numerical output from the KDE toolbox look like? Is it empty, or does it contain only NaNs? How do you call it? Is there an error message? Again: How do your data look like, how many data points, how many dimensions? Can you post a link to the data set? – A. Donda Dec 27 '13 at 14:39

0 Answers0