Problem with "Could not find a proper state to initialise. Perhaps try another seed."

[dachao liang]

Dear BEAST user,

I using BEAST2.3.2 to run the zv_ns5_131_yfvref_mrca3.xml to calculate the TMRCA. But the BEAST2.3.2 generate many error massages unfortunately.I try to change the upper boundary as other post said.It fail to run with the same messages. Would you like to give me some idea about this?Than you.

Start likelihood: -Infinity after 10 initialisation attempts

P(posterior) = -Infinity (was NaN)  **

        P(prior) = -Infinity (was NaN)  **

                P(CoalescentConstant.t:zv_ns5_131_yfvref_1) = -319.56888921383205 (was NaN)  **

                P(ClockPrior.c:zv_ns5_131_yfvref_1) = -1.4620820845188824 (was NaN)  **

                P(GammaShapePrior.s:zv_ns5_131_yfvref_1) = -1.0 (was NaN)  **

                P(GammaShapePrior.s:zv_ns5_131_yfvref_2) = -1.0 (was NaN)  **

                P(GammaShapePrior.s:zv_ns5_131_yfvref_3) = -1.0 (was NaN)  **

                P(kappa1Prior.s:zv_ns5_131_yfvref_1) = -1.8653600339742873 (was NaN)  **

                P(kappa1Prior.s:zv_ns5_131_yfvref_2) = -1.8653600339742873 (was NaN)  **

                P(kappa1Prior.s:zv_ns5_131_yfvref_3) = -1.8653600339742873 (was NaN)  **

                P(kappa2Prior.s:zv_ns5_131_yfvref_1) = -1.8653600339742873 (was NaN)  **

                P(kappa2Prior.s:zv_ns5_131_yfvref_2) = -1.8653600339742873 (was NaN)  **

                P(kappa2Prior.s:zv_ns5_131_yfvref_3) = -1.8653600339742873 (was NaN)  **

                P(PopSizePrior.t:zv_ns5_131_yfvref_1) = -1.4925080392996117 (was NaN)  **

                P(YFV_america.prior) = -Infinity (was NaN)  **

        P(likelihood) = NaN (was NaN)  **

                P(treeLikelihood.zv_ns5_131_yfvref_1) = NaN (was NaN)  **

                P(treeLikelihood.zv_ns5_131_yfvref_2) = NaN (was NaN)  **

                P(treeLikelihood.zv_ns5_131_yfvref_3) = NaN (was NaN)  **

java.lang.Exception: Could not find a proper state to initialise. Perhaps try another seed.

        at beast.core.MCMC.run(Unknown Source)

        at beast.app.BeastMCMC.run(Unknown Source)

        at beast.app.beastapp.BeastMain.<init>(Unknown Source)

        at beast.app.beastapp.BeastMain.main(Unknown Source)

        at beast.app.beastapp.BeastLauncher.main(Unknown Source)

Best regards,

Victor

[dachao liang]

Dear BEAST user,

For this xml file,I used tip date and internal calibration together to calculate TMRCAs.But the tutorial didn't use tip date.Would it caused the error massages?Thank you.

Regards,

Victor

在 2016年4月15日星期五 UTC+8下午7:38:30，dachao liang写道：

[Susana Freitas]

Dear all,
I am having the same exact error with beast v2.4.2. I am running a set of 40 individuals and 30 markers for 500M in StarBEAST2. I tried both strict clock and uncorrelated Lognormal (with some priors to set monophyletic groups, with uniform distribution, and the tMRCA, with Log normal distribution). For both xmls I get this error message

Start likelihood: -Infinity after 10 initialisation attempts

Fatal exception: Could not find a proper state to initialise. Perhaps try another seed.

/

Programmer error: input rule should be REQUIRED for this Input constructor (populationmodel)
(...)
P(posterior) = -Infinity (was -Infinity)
        P(speciescoalescent) = 1667.416414294123 (was 1667.416414294123)
        P(prior) = -Infinity (was -Infinity)
                P(CalibratedYuleModel.t:Species) = -Infinity (was -Infinity)
(...)
P(chlo_der.prior) = NaN (was NaN)  **
                P(ingroup.prior) = NaN (was NaN)  **
                P(mix2_par.prior) = NaN (was NaN)  **
                P(outgroup.prior) = NaN (was NaN)  **
                P(por.prior) = NaN (was NaN)  **
                P(pra.prior) = NaN (was NaN)  **
                P(rud.prior) = NaN (was NaN)  **
                P(val_group.prior) = NaN (was NaN)  **

        P(likelihood) = NaN (was NaN)  **

                P(treeLikelihood.Locus_1612_3025) = NaN (was NaN)  **
                P(treeLikelihood.Locus_3686_2810) = NaN (was NaN)  **
                P(treeLikelihood.Locus_4045_3049) = NaN (was NaN)  **
                P(treeLikelihood.Locus_4492_929) = NaN (was NaN)  **
                P(treeLikelihood.Locus_4803_3740) = NaN (was NaN)  **
                P(treeLikelihood.Locus_5164_1549) = NaN (was NaN)  **
                P(treeLikelihood.Locus_5249_3291) = NaN (was NaN)  **

When I removed the priors on the monophyletic groups the program started running with no issues. The only error that remained was the "Programmer error: input rule should be REQUIRED for this Input constructor (populationmodel)".
Do you know why this happened? I thought that if we knew already that some groups were monophyletic that would help the analysis, and not disrupt it.

Thanks in advance!
Susana

[Huw A. Ogilvie]

Hi Susana,

Could you upload one of the XML files that won't initialize?

[Rebecca Harris]

Did you ever find a solution to this issue? I am running into the same problem...

[Nick Matzke]

Typically whatever produces Infinity/NaN first is what is causing the problem, so it's probably P(CalibratedYuleModel.t:Species).  

I have a vague memory that I had a similar problem with the tree prior once, and solved it by putting in more plausible starting rates for speciation/extinction (or netDiversification etc.).  I may have also hard-coded a plausible starting tree.

Cheers, Nick 

--
You received this message because you are subscribed to the Google Groups "beast-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to beast-users+unsubscribe@googlegroups.com.
To post to this group, send email to beast...@googlegroups.com.
Visit this group at https://groups.google.com/group/beast-users.
For more options, visit https://groups.google.com/d/optout.

[Huw A. Ogilvie]

Hi Rebecca,

Could you upload the XML file that won't initialize?