quarnetGoFtest! Error

[Yasuto ISHII]

Dear Solis-Lemus and Ane,

I ran quarnetGoFtest! function to get the support for my networks,

but it terminated with an error.

My data are ddRAD-seq data,

so I calculated CF tables via SNPs2CF R package.

The raw CF table of SNPs2CF cannot import, so

1) I removed 4 columns named *_lo and *_hi

2) I selected one row (CFs) per one four-taxon set randomly

The modified CF table seems to be the same as the one shown in the tutorial

(https://cecileane.github.io/QuartetNetworkGoodnessFit.jl/stable/man/gof/#Goodness-of-fit-of-a-candidate-network).

The error (error.txt), the CF table (csv file) and the script (quarnetGoFtest.jl) to run quarnetGoFtest! function are attached.

Sincerely,

Yasuto

[Yasuto ISHII]

Hi all,

It seems to run correctly

after  I removed the integers after the right parentheses from .network or .out file.

Alternatively, I faced a warning.

The warning is:

┌ Warning: The simulated z values are far from 0 and they shouldn't:
│ with a mean of 1.8119 and a standard deviation of 4.3311.
│ The network might be in a form that causes a bug in the hybrid-Lambda simulator
│ (perhaps related to this: https://github.com/hybridLambda/hybrid-Lambda/issues/36).
└ @ QuartetNetworkGoodnessFit ~/.julia/packages/QuartetNetworkGoodnessFit/NhwsJ/src/quarnetGoF.jl:253

How can I determine whether it is a bug or not?

Best regards,

Yasuto

2022年7月16日土曜日 1:19:21 UTC+9 Yasuto ISHII:

[Cécile Ané]

Hi Yasuto,

Version 0.15.0 of PhyloNetworks should have fixed the issue you encountered with HybridLambda, which bugs if the phylogeny has bootstrap values as internal node names.

For the warning: this warning can happen if you tried a short trial run with a handful of simulations only (like nsim=3 in the example). But if you get this warning with more simulations, then this is a serious issue. I think it's a signal that HybridLambda has a hidden bug on your phylogeny, unfortunately.

Cecile.

[Yasuto ISHII]

Dear Cecile,

I ran the function with nsim = 1000 as recommended in the tutorial.

The script is:

res = quarnetGoFtest!(net, qCF, false; seed=1, nsim=1000)

I will quit using the test.

Then, I am wondering how I can determine the optimal No of reticulation?

Some authors (eg., Olave et al. 2020; Karimi et al. 2020) determined the number whose loglikelihood is lowest;

others (eg., Esquerre et al. 2022) adopted the number that improves loglikelihood the most.

My reticulation number-loglikelihood plot is attached.

Best regards,

Yasuto

2022年7月17日日曜日 23:43:43 UTC+9 Cécile Ané: