[Mesquitelist] Likelihood reconstruction of changes: how to generate a table?
wmaddisn at interchange.ubc.ca
Thu Sep 11 22:38:10 PDT 2008
With likelihood it's not clear to me exactly how you'd do this.
Likelihood doesn't measure the probability of a state's occurrence at
a node, but rather the probability of observing the data given the
state's occurrence at the node. We've never tried to calculate the
likelihood of a change -- i.e. the probability of the data given a
state change on a branch. An interesting idea, but I'm not aware
that anyone has used the calculation. (It could be done relatively
easily for categorical characters as a sum of node state
likelihoods.) Would you be interested in the sum over all characters
of the likelihoods of change on a branch?
Another option would be to use stochastic character mapping to
estimate distributions of total numbers of changes on a branch. This
would involve a new module to sum over realizations.
What exactly would you want?
At 4:56 PM -0400 8.8.2008, David Fitch wrote:
>Dear Mesquite Users/Designers,
>For a matrix of discrete binary characters, how does one output a
>table that details the likelihood reconstruction of character-state
>CHANGES ("losses" and "gains") in each character and each branch?
>I see that Mesquite can output a table of the probabilities of
>HAVING a particular state at each branch for each character
>(Analysis > TraceAllCharacters). And I see that Mesquite can output
>a table of changes occurring across a (or many) tree(s) for ONE
>particular character at a time (Taxa&Trees >
>SummarizeStateChangesOverTrees). However, I cannot figure out how
>to output a table of CHANGES for EACH branch for EACH character.
>For example, each character could be represented by a row in the table.
>Each branch is represented by a pair of columns:
> the 1st column with the number of gains,
> the 2nd column with the number of losses.
>Presumably, the "number of gains" or "number of losses" would not
>necessarily be integers, because the probability of a particular
>change may not be 1.0 or 0.0 for a particular branch (unless the
>decision threshold is applied; e.g. >50% probability for a gain
>would be scored as a definite gain for that branch).
>Any ideas would be very welcome!
>~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ~ ~ - -
>~ David H. A. Fitch ~ \ / / /
>~ Associate Professor ~ \/ / /
>~ Department of Biology ~ \ / /
>~ New York University ~  / /
>~ Main Building, Room 1009 ~ \ / /
>~ 100 Washington Square East ~ \/ /
>~ New York, NY 10003 ~ \ /
>~ U S A ~ \/
>~ Tel.: (212) 998-8254 ~ \
>~ Fax: (212) 995-4015 ~ \
>~ e-mail: <mailto:david.fitch at nyu.edu>david.fitch at nyu.edu ~ \
>Mesquitelist mailing list
>Mesquitelist at mesquiteproject.org
Professor and Canada Research Chair
Depts. of Zoology and Botany and
Biodiversity Research Centre
Beaty Biodiversity Museum
6270 University Boulevard
University of British Columbia
Vancouver, BC V6T 1Z4 Canada
email: wmaddisn at interchange.ubc.ca FAX: +1 604 822-2416
Tree of Life: http://tolweb.org
Beaty Biodiversity Museum: http://beatymuseum.ubc.ca
Home page: http://salticidae.org/wpm/home.html
More information about the Mesquitelist