8
$\begingroup$

What non-/semiparametric methods to estimate a probability density from a data sample are you using ?

(Please do not include more than one method per answer)

$\endgroup$
  • $\begingroup$ your experience -- has run on 1k / 100k points, in 1d / 2d / 3d -- would also be useful. $\endgroup$ – denis Oct 21 '10 at 14:16
5
$\begingroup$

Dirichlet Process mixture models can be very flexible nonparametric Bayesian approach for density modeling, and can also be used as building blocks in more complex models. They are essentially an infinite generalization of parametric Gaussian mixture models and don't require specifying in advance the number of components in the mixture.

$\endgroup$
5
$\begingroup$

Gaussian Processes can also be another nonparametric Bayesian approach for density estimation. See this Gaussian Process Density Sampler paper.

$\endgroup$
3
$\begingroup$

I use Silverman's Adaptive Kernel Density estimator. see e.g akj help page

$\endgroup$
2
$\begingroup$

Half-space depth a.k.a. bag-plots.

http://www.r-project.org/user-2006/Slides/Mizera.pdf

$\endgroup$
1
$\begingroup$

A nice short paper by Jose Bernardo here gives a useful Bayesian method to estimate a density. But as with most things Bayesian, the computational cost must be paid for this method.

$\endgroup$

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.