Increase ESS Values for (prior, posterior, likelihood, ORCucldsigma, ORCsigma)

[Mohammed Houta]

Dears,

Greetings,

Kindly find the attached ESS for my last run on beast 2.6.3. 

Settings on beauti were as follows:

1. Site Model: bModeltest +estimate mutation rate. 

2. Clock model: optimised relaxed clock + estimate clock rate 

3. Prior: Coalescent bayesian skyline+ mrca [Var2 + lognromal+ monophyletic+ M  value (1998.0)+ M lower(1930.0)+  M upper (2000.0) + check Mean in real space ]. 

4. MCMC: 100 M + pre burnin (100K) + tracelog (1K)+ screenlog (1K)+ treelog(1K). 

5. Mode: automatic set clock rate is checked + Automatic set fix mean sub. Rate flag is checked. 

6. Operators: tree full s1 scales all internal nodes (scale factor: 0.81)+ Delta exchange BMT mutation rate ( scale factor: 10417.88). Those values were suggested by beast for the previous run. 

What should I do to increase the ESS results? 

Increase weight? Increase Scale Factors? Increase MCMC length? Increase Sampling intervals? 

Please suggest the most appropiate action to increase ESS. 

Thanks in advance.

M. Houta

[Tzu-Hao Kuo]

Hi, 

Which methods suggested by the official tutorials have you tried?

Tzu-Hao

--
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 view this discussion on the web visit https://groups.google.com/d/msgid/beast-users/CAPYWc7mDqiXb47f1JnDDtYnsQ2PzTbGvc6Eb92bwGTGxuOUZYA%40mail.gmail.com.

[HS]

Hi,

I would suggest to try a run without the estimate of the clock rate. I assume you have serial sample, so it will calibrate the clock, you don't need to use model with estimating it.

Best,

Hovhannes

[Mohammed Houta]

No official method, It's a trial and error. 

M. Houta

To view this discussion on the web visit https://groups.google.com/d/msgid/beast-users/CAMEtALWfb0%2Bvq0XyVjONeUyXPkGRPAV31MZXm0n3KidC7V7%2BRQ%40mail.gmail.com.

[Mohammed Houta]

I will try it, But I hope that you would help me with MRCA tuning because the numbers which I used are in recent years (1998, 1930, 2000) not in Million years.  Kindly find the attached pics 

M. Houta. 

--

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 view this discussion on the web visit https://groups.google.com/d/msgid/beast-users/cf5be255-da66-40fc-b564-d88a2c1342fen%40googlegroups.com.

[HS]

I think you can use tip dates leaving default priors for other clock-related parameters.

Cheers,

Hovhannes

[Mohammed Houta]

Dear Dr. Hovhannes

Greetigs

Could I preburnin 70-80% of trees file in annotating process to avoid GC over head error? 

FYI, the remaining trees for annotation would be 20000 or 10000 respectively. 

Thanks in advance

M. Houta

To view this discussion on the web visit https://groups.google.com/d/msgid/beast-users/f7816a9d-b375-49db-96c5-70950d2d4548n%40googlegroups.com.

[HS]

Hi M. Houta,

I would not be satisfied if only the remaining 20% chains are OK. If 80% of chains are still fluctuating, it may be indicative of an improper or insufficiently long analysis.

Also, you don't need to sample so frequently. In my experience it gives nothing if the analysis itself is not satisfactory. For the whole analysis 10000 trees are well enough.

Best,

Hovhannes

[Carlo Pacioni]

Hi M. Houta,

you mentioned in a earlier email that you had an issue with the unit of the clock, which seems to be in million years while you thought you used years. BEAST interprets node calibration in years before the most recent sample, but it seems that you have entered the calibration as dates (M  value (1998.0)+ M lower(1930.0)+  M upper (2000.0)). If, for example, your most recent samples is in 2021 and you want the node to have a mean age in 1998, your M value should be 23. Hope this makes sense.

Cheers,

carlo

To view this discussion on the web visit https://groups.google.com/d/msgid/beast-users/0fd72e10-3b9a-461d-855f-6cb90d8f1d59n%40googlegroups.com.