Can run with more particles?

[richelbilderbeek]

When doing a run with 10 particles, results look the same.

Is this a bug?

[richelbilderbeek]

• [x] [edit: this is OK] Check that epsilon is low. If not, flawed in razzo parameter setting

richel@sonic:~/data/razzo_project_20190905/data/0.2-0-1-0.1/1$ Rscript -e 'readRDS("parameters.RDa")'

[...]

$pir_params$experiments[[4]]$est_evidence_mcmc$epsilon
[1] 1e-12

• [x] [edit: cannot check this] Check if BEAST2 .xml files differ. If not, bug is found in babette/mcbette/pirouette/razzo

Cannot check this, as the marginal likelihood estimation is done in the backend. I consider storing these logs as well, after trying the next steps.

• [x] [edit: cannot check this] When running BEAST2, observe it output (when it reads the parameters) differ. If not, bug is found in BEAST2

Cannot check this, as the marginal likelihood estimation is done in the backend. I consider storing these logs as well, after trying the next steps.

• [ ] Check for a longer/shorter DNA sequence length: the unchanged uncertainties may reflect there is not enough genetic information to decrease the uncertainty

[richelbilderbeek]

Started a run with a DNA alignment length of 100. I predict the marginal likelihood estimates will be larger.

[richelbilderbeek]

Gotcha! There is no argument for the number of particles in est_marg_lik!

[richelbilderbeek]

Escalated this Issue to fixing the bug, #280