The initial model is invalid because state has a zero probablity, Beast

[pawan jayaswal]

Dear BEAST user,

 I got error like "The initial model is invalid because state has a zero probability"  any one would like to suggest me  how to short out this error.

  I have used following parameter  to generate .XML file using BEAUTi.

1.Taxa:7 (All r having different fossils information:Min120myrs and maxm 1200myrs)

2.Sites: HKY Model, Site Heterogeneity: Gamma, partition into codon positions: 3 partititons:position1,2,3

3.Clocks:Lognormal relaxed clock(Uncorelated)

4.Trees: Speciation:Birth-Death process, Random Starting tree

5.States:Reconstruct states at ancestor and others checked

6.Priors:

 tmrca:::::For rooted (Gamma distribution, Shape 2.0, Scale 2.0 offset 1190),

 tmrca:::::For others (Exponential, Mean 5-12, offset : fixed a/cding to the fossils nformation)

 CP1,2.MU::::Exponential, initial value 1.0, Mean 0.5, offset 0, CP3.mu, mean5.0

 UCLD.mean;;;Expo, Init val:1.0, Mean 0.05, offset 0.0

TreeModel.rootheight: gamma, Initial:1210.0, Shape:3.5, Scale 8.0, offset :1200

birth Death.meangrowthrate: EXPONENTIAL, iNITIAL VALUE 1250, Mean value :0.15, OFFSET:0.0

After generating XML file with above mention parameter I got an error like  "The initial model is invalid because state has a zero probability, If the log likelihood of the tree is -Inf, his may be because the initial, random tree so large that it has an extremly bad likekihood which is being rounded to zero".

I have checked and changed my existing parameter but could not get any positive solution for this.

I will be thankful, if any body can suggest me how to shortout this issue. Error file is attached with this..

Pawan kumar jayaswal

[Krzysztof M. Kozak]

Dear Pawan,

Classic! You'll find lots on this error posted on the forum already. It usually means that your priors are in conflict. Look at the output and find the parameter that has log likelihood equal to -Inf. Looks like three of your tmrca values are the problem - maybe they don't fit the starting tree?

Also, did you try a simpler combination of models first? You have many complicated things going on (UCLD clock, birth-death tree prior, ancestral states): it may be easier to start with the simplest settings (strict clock, yule tree, etc) and build up a more complicated scheme, if you haven't done so already.

Cheers,

[pawan jayaswal]

Dear Kozak,

   Thank you very much for your reply. I have changed  tmrca prior (Av,Fi,Amphibia)  from exponential to uniform/Normal and after that BEAST software run well but when generated log file passed from the  Tracer software Av, Fi and amphibia has shown blank (-) ESS value.

 

mean	0
stderr of mean	n/a
stdev	n/a
variance	n/a
median	n/a
mode	n/a
geometric mean	n/a
95% HPD Interval	n/a
auto-correlation time (ACT)	n/a
effective sample size (ESS)	n/a

Would you like to make a comment on this.

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!