SFS noise model - Githubissues

harrispopgen / mushi

[mu]tation [s]pectrum [h]istory [i]nference

https://harrispopgen.github.io/mushi/

MIT License

24 stars 5 forks source link

SFS noise model #10

Closed wsdewitt closed 5 years ago

wsdewitt commented 5 years ago

We're using a Poisson random field approximation to compute Pr(sfs | xi). Yun's papers (i.e. fastNeutrino) use a Kullback-Liebler divergence D(sfs || xi). Why? It seems this would miss the fact that higher components have more noise.

wsdewitt commented 5 years ago

PRFFTW