NaN/ Infinite Marginal Likelihood Estimate in SNAPP/BFD*

[tkch...@email.uark.edu]

I am running BFD* species delimitation using SNAPP and Beast_2.2.1 (I couldn't get it to work with the more recent Beast... Spits out a bunch of indecipherable java errors) and I am getting in certain runs an infinite likelihood in the std output, which causes a NaN for the final marginal likelihood calculation. 

I am new to Beast, but I saw on some other posts (not pertaining to SNAPP/BFD in particular..) that a parameter might end up at a extreme value and cause the likelihood to be calculated incorrectly. Is there a way to prevent/fix this, or is there a way for me to take the marginal likelihoods from the individual "steps" and calculate the final corrected marginal likelihood for the entire model? I've pasted the results from one of my runs below (which was 50k pre-burnin and chainLength=200k, with everything else set as in the BFD tutorial), and as you can see a single step ended up with a weird likelihood, but I have the marginal likelihoods for the rest. If it is a legitimate bug causing the 1 step going weird, how could I go about using the estimates from the other 47 steps?

Thanks you! Any help is much appreciated...

Step        theta         likelihood   contribution ESS

0            1            -9875.485    -683.2539    492.1718

1            0.9308       ∞            �            �

2            0.8651       -9881.1415   -617.0052    299.434

3            0.8026       -9884.6487   -585.367     386.5781

4            0.7434       -9888.405    -554.6711    265.5121

5            0.6873       -9892.7571   -524.9368    355.0735

6            0.6343       -9896.1421   -496.0646    224.1079

7            0.5842       -9900.4625   -468.1639    274.6605

8            0.5369       -9905.3879   -441.2018    312.3482

9            0.4924       -9909.3947   -415.0866    452.3919

10           0.4505       -9913.8068   -389.8985    408.0754

11           0.4112       -9918.758    -365.6141    481.9881

12           0.3743       -9923.4008   -342.1997    585.1401

13           0.3398       -9928.6706   -319.6839    405.8054

14           0.3076       -9933.7173   -298.0104    680.4721

15           0.2777       -9939.6619   -277.2363    551.8107

16           0.2498       -9945.9494   -257.2896    564.8967

17           0.2239       -9952.3898   -238.1945    684.692

18           0.2          -9958.8612   -219.9377    801

19           0.1779       -9966.5595   -202.5042    801

20           0.1576       -9974.4838   -185.8901    545.7912

21           0.139        -9982.55     -170.0723    700.6719

22           0.1219       -9991.4183   -155.059     692.2584

23           0.1064       -9999.1669   -140.8064    558.0086

24           0.0923       -10007.2435  -127.3315    635.5343

25           0.0796       -10018.2701  -114.6444    488.8382

26           0.0682       -10027.0119  -102.6872    556.5758

27           0.058        -10037.6866  -91.4797     713.9925

28           0.0488       -10049.5361  -81.0052     545.9281

29           0.0408       -10166.7018  -72.2712     3.1833

30           0.0337       -10093.6102  -62.4352     20.4588

31           0.0275       -10300.0133  -55.6603     4.8865

32           0.0222       -10625.9439  -48.6517     32.686

33           0.0177       -10581.648   -41.0994     8.9432

34           0.0138       -10723.8416  -34.682      360.6881

35           0.0106       -10760.9678  -28.6534     304.4289

36           0.0079       -10806.0258  -23.2654     795.6527

37           0.0058       -10826.0791  -18.4554     758.3903

38           0.004        -10854.2459  -14.2761     801

39           0.0027       -10878.9842  -10.6885     801

40           0.0018       -10890.1844  -7.6666      801

41           0.001        -10902.658   -5.2033      801

42           0.0006       -10910.8742  -3.2667      770.7563

43           0.0003       -10906.5523  -1.824       801

44           0.0001       -10907.1812  -0.8402      738.2622

45           0            -10902.8902  -0.2642      801

46           0            -10910.3519  -0.0291      538.0849

47           0            -10903.2291  0            703.062

marginal L estimate = NaN

[Remco Bouckaert]

You could rerun the analysis for step 1 — the one with the infinite likelihoods — and run the PathSamplaAnalyser again to calculate the marginal likelihoods.

To rerun step 1, remove the log and tree files from the directory “step1” inside the root directory specified in your XML file. Then run BEAST on beast.xml in step1.

Cheers, Remco

--
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...@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.

[tkch...@email.uark.edu]

Thanks, will give it a try