Stepping Stone in BEAST 2

[Sarah N.]

Hi Beast - users,

is it possible to run Stepping Stone in BEAST 2?

I only find tutorials and skripts for Path Sampling in Beast 2, or Stepping Stone in Beast1.


Thank you for your replies!

Sarah

[Huw A. Ogilvie]

The "path sampling" available in BEAST2's MODEL_SELECTION is actually stepping-stone sampling. It is a straightforward implementation of the stepping-stone method described in Xie et al. Path sampling and stepping-stone sampling are similar - stepping-stone is slightly more efficient, but given enough steps either should be very accurate.

[Arafat Rahman]

Dear Ogilvie,

I'm similarly confused here. I've gone through the BEAST2's model selection tutorial and able to estimate PS score. But I don't find any way to estimate SS score. Can you help to clear this confusion?

BR,
Arafat

[Remco Bouckaert]

Hi Arafat,

If set up the analysis and left alpha to 0.3 (which is the default value), you actually calculated the stepping stone estimate with the (confusingly named) PathSampleAnalyser.

If you want to calculate the path sampling estimate instead of the stepping stone estimate, you should set alpha to 0. Be aware that the alpha value in the analysis run with BEAST and in the PathSampleAnalyser should be the same.

Hope this clarifies things,

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.

[Arafat Rahman]

Dear Remco,

Thanks for your clarification. It's clear now.

Best Regards,
Arafat

[xumei...@gmail.com]

Hi Remco,

I follow the "Step 11bEditing the XML file for marginal likelihood estimation" in BFD_-tutorial to set my XML file to run the stepping stone analysis with BEAST2,but I got error,details as :

Error 1017 parsing the xml input file

Class could not be found. Did you mean snap.operators.GammaMover?

Error detected about here:

  <beast>

      <run spec='beast.inference.PathSampler'>

I wonder is it right to run stepping stone analysis with BEAST2 in this way  ?or is there something else I need to do?

thanks,Jessica

在 2017年4月27日星期四 UTC+8上午5:25:10，Remco Bouckaert写道：

[nam...@cornell.edu]

Hi everyone,

I'm experiencing the same problem as Jessica above me. I've followed the directions to set up BFD* with SNAPP by editing the XML file, but get the same error:

Error 1017 parsing the xml input file

Class could not be found. Did you mean snap.operators.GammaMover?

Error detected about here:

  <beast>

      <run spec='beast.inference.PathSampler'>

Oddly enough, I had BFD* + SNAPP working on a different data set a couple years ago, so I'm not sure what's causing this error as now it won't work on files that used to run fine. Any thoughts?

[Remco Bouckaert]

Hi Jessica,

Thanks for the XML file — it runs fine on my computer.

On the computer that you run the analysis, can you make sure that the model-selection package installed?

Cheers,

Remco

[Josh Banta]

Hi -- I was having the same problem. You first need to open BEAUTI and then go to "File" ==> "Manage Packages." Then you need to install the "MODEL_SELECTION" package.

Then you can run your XML file in BEAST. Your problems will be solved!

Josh

[Fushi Ke]

Hi Remco,

I am running path sampling followed your suggestions by setting alpha to 0. But I always get errors when all 48 steps finished:

#############################################

....

marginalLs[42 ] = -11081.2701358706
marginalLs[43 ] = -11096.121746759627
marginalLs[44 ] = -11118.263020873974
marginalLs[45 ] = -11155.197286286615
marginalLs[46 ] = -11292.692022620597
marginalLs[47 ] = -12166.590722255836

Step        theta         likelihood   contribution ESS
0            1            -10954.2576  0            194.3244    
1            Fatal exception: Cumulative probability function returned NaN for argument 0.979 p = 0.979

###############

I got normal ESS and likelihood when alpha was set to 0.3.

Can you give me some suggestions about this?

Best regards,

Fushi

在 2017年4月27日星期四 UTC+8上午5:25:10，Remco Bouckaert写道：

Hi Arafat,

If set up the analysis and left alpha to 0.3 (which is the default value), you actually calculated the stepping stone estimate with the (confusingly named) PathSampleAnalyser.

If you want to calculate the path sampling estimate instead of the stepping stone estimate, you should set alpha to 0. Be aware that the alpha value in the analysis run with BEAST and in the PathSampleAnalyser should be the same.

Hope this clarifies things,

Remco

On 24/04/2017, at 3:39 PM, Arafat Rahman <ac.a...@gmail.com> wrote:

Dear Ogilvie,

I'm similarly confused here. I've gone through the BEAST2's model selection tutorial and able to estimate PS score. But I don't find any way to estimate SS score. Can you help to clear this confusion?

BR,
Arafat

On Friday, July 8, 2016 at 6:16:54 PM UTC+6, Huw A. Ogilvie wrote:

The "path sampling" available in BEAST2's MODEL_SELECTION is actually stepping-stone sampling. It is a straightforward implementation of the stepping-stone method described in Xie et al. Path sampling and stepping-stone sampling are similar - stepping-stone is slightly more efficient, but given enough steps either should be very accurate.

On Friday, 8 July 2016 19:46:43 UTC+10, Sarah N. wrote:

Hi Beast - users,

is it possible to run Stepping Stone in BEAST 2?

I only find tutorials and skripts for Path Sampling in Beast 2, or Stepping Stone in Beast1.


Thank you for your replies!

Sarah

--
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...@googlegroups.com.

[Remco Bouckaert]

Hi Sushi,

This turned out to be caused by a bug in the model selection package. It is fixed now and v1.5.3 of the package should not have this problem any more.

Cheers,

Remco

To unsubscribe from this group and stop receiving emails from it, send an email to beast-users...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/beast-users/3d577973-ba24-4fec-b732-6e665532fe5c%40googlegroups.com.
<spedem.intergene.bi.1.005.10k.clade1.model1.00.xml><o.model1.00.clade1>