HyPhy message board - likelihood per position/partition

	Welcome, Guest. Please Login

Home

Help

HyPhy message board › Methodology Questions › How to › likelihood per position/partition

(Moderators: Sergei, Simon)

‹ Previous Topic | Next Topic ›

Pages: 1

Send Topic

likelihood per position/partition (Read 2752 times)

gustavo YaBB Newbies Offline I love YaBB 1G - SP1! Posts: 12	likelihood per position/partition Aug 12^th, 2004 at 3:23pm How can I print the likelihood for each position of the alignment or for each partition? Thanks very much in advance, gustavo
Back to top	IP Logged

Sergei

YaBB Administrator

Offline

Datamonkeys are forever...

Posts: 1658
UCSD
Gender: male

Re: likelihood per position/partition
Reply #1 - Aug 12^th, 2004 at 5:13pm

Dear Gustavo,

The basic idea is to use a 'ConstructCategoryMatrix' command to tabulate site-by-site likelihood. This gets a little trickier is there are category (rate variation variables) involved, because the command outputs a vector of marginals for each site.

Multimedia File Viewing and Clickable Links are available for Registered Members only!! You need to

Login

points to an example file (it's self-contained NEXUS) - look at the source of the example in the 'HYPHY' block to see how this works. Just open this a a regular batch file to execute the example.

In the GUI version (Mac and Windows), you can use 'Likelihood->Category Processor' menu from the parameter table window (which you can get to by opening the Object Inspector, finding the tree in the current analysis, opening its window and they choosing 'Tree->Show Parameters in Table' menu) to display a table with site-by-site likelihoods.

Sergei

Associate Professor
Division of Infectious Diseases
Division of Biomedical Informatics
School of Medicine
University of California San Diego

WWW

IP Logged

Sergei

YaBB Administrator

Offline

Datamonkeys are forever...

Posts: 1658
UCSD
Gender: male

Re: likelihood per position/partition
Reply #2 - May 3^rd, 2006 at 1:19pm

In case there are rate categories...

Assume for a second that you have a single category variable 'c' with D classes. If you call
ConstructCategoryMatrix on a likelihood function with 'c' in the model and N sites, you'll get an DxN matrix, where the entry in row i and column j tells you the likelihood of site j, given that 'c' takes value from its i-th class.

To compute the likelihood of a site then is simply a matter of computing expected values. You can do this as follows (replace variables as appropriate).

Code:

/* this needs HyPhy from 2006 for row indexing tricks */

GetInformation (catInfo, c);
ConstructCategoryMatrix (cmx, lf, COMPLETE);
/* extract the second row from catInfo - the weights for each class */
weights = catInfo[1][-1];
/* compute the expectations; these will be likelihoods; NOT log-likelihoods */
lkBySite = weights *cmx;
/* compute log-L */
logLBySite = Log (lkBySite);

« Last Edit: May 4^th, 2006 at 11:59am by Sergei »

Associate Professor
Division of Infectious Diseases
Division of Biomedical Informatics
School of Medicine
University of California San Diego

WWW

IP Logged

Pages: 1

Send Topic

‹ Previous Topic | Next Topic ›

« Home

‹ Board

Top of this page