SEVERE: The initial model is invalid because state has a zero probability

[Chris Shields]

Hello-

I am very new to BEAST, and I am working with a large data set (260
sequences). I am trying to date divergences between taxa, but always
get the following error.

The individual components of the posterior are as follows:
The initial posterior is zero: prior=(JeffreysPriorLikelihood=0.0000,
JeffreysPriorLikelihood=-17.4264, coalescent=-4513.4524), likelihood=
(treeLikelihood=-Inf)

I read on here about using a user generated tree, so I used r8s to get
a tree with years, but when I run BEAST I still get the same error.
My tree appears to load properly, but I cannot get the analysis to
run. Please send me any suggestions. I can provide more details if
they will help. Thanks in advance.
~Chris

[Couvreur, Thomas]

Hi Chris,

When I have this error there is something wrong with the taxa sets you create in BEAUTi and the monophyly command. If you forget to include one species in the monophyletic taxa set, then your starting tree will not reflect that. Carefully go through your taxa sets and see if they are present in the starting tree.

 

One other possibility is the that the starting tree itself is quite bad... I usually use a ML tree from RAxML (which is a fast and quick ML way to get a very good starting tree in most cases).

Hope this helps.

Thomas

---

From: beast...@googlegroups.com on behalf of Chris Shields
Sent: Sat 2/21/2009 3:36 PM
To: beast-users
Subject: SEVERE: The initial model is invalid because state has a zero probability