turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

Find a Community

- Home
- /
- SAS Programming
- /
- SAS Procedures
- /
- How can I specify a prior in proc fmm using the Ba...

Topic Options

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

05-04-2016 07:52 AM

I want to estimate a univariate Gaussian mixture model with two components using proc fmm. With respect to the Bayes statement, I am assuming a triangular distribution for one parameter. Does anybody know how to implement it? I checked proc mcmc and there it is possible to construct your own prior via the general function (https://support.sas.com/documentation/cdl/en/statug/63033/HTML/default/viewer.htm#statug_mcmc_sect02...

Many thanks in advance.

Accepted Solutions

Solution

05-16-2016
12:01 PM

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

05-04-2016 10:37 AM

I don't think that FMM supports general priors. I only see the "usual suspects" in the doc: beta, gama, Dirichlet, etc

However, there is an easy way to create a beta distribution that kinda-sorta looks like a triangular distribution, in the sense that it has support is on [a,b] and it has a mode at c. It is sometimes known as the PERT distribution. See the article "That distribution is quite PERT!" to learn how to choose the parameters of the beta distribution. Choosing a PERT (via beta) distribution might capture the spirit of your prior.

All Replies

Solution

05-16-2016
12:01 PM

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

05-04-2016 10:37 AM

I don't think that FMM supports general priors. I only see the "usual suspects" in the doc: beta, gama, Dirichlet, etc

However, there is an easy way to create a beta distribution that kinda-sorta looks like a triangular distribution, in the sense that it has support is on [a,b] and it has a mode at c. It is sometimes known as the PERT distribution. See the article "That distribution is quite PERT!" to learn how to choose the parameters of the beta distribution. Choosing a PERT (via beta) distribution might capture the spirit of your prior.