LD regularization missing

[soisa001]

Hello,

I have tried to use phlash and it generally runs fine, but the output size histories (when simulated with msprime/stdpopsim) do not correspond well in terms of LD decay to the input test data. The SFS does fit fairly well, so it is promising in this metric. The paper describes the LD regularization, but it does not seem to be implemented in model.py log_density function. Am I missing something, can I make the run fit better by LD decay, or will this be implemented in a later version?

Thanks,
Natthapon

[terhorst]

Hi, sorry for the delayed response. This code currently lives on the fit_refactor branch, see the phlash.ld submodule. I've yet to merge it into the main tree. LD regularization was added in revision and is not as thoroughly tested as the main inference code.