And will this then allow us to use the method as a Gibbs sampler for MCMC, as the Siepel group do with their lineage re-threading method? If so, it might have the potential to be much more powerful than we first imagined.
Many bridges to be solved before getting there, though. Must also look into why ARGweaver discretises time.
We will need someone with a good stats head to think about all this. Gil?
Copied from original issue: mcveanlab/treeseq-inference#3
From @hyanwong on December 5, 2016 9:7
And will this then allow us to use the method as a Gibbs sampler for MCMC, as the Siepel group do with their lineage re-threading method? If so, it might have the potential to be much more powerful than we first imagined.
Many bridges to be solved before getting there, though. Must also look into why ARGweaver discretises time.
We will need someone with a good stats head to think about all this. Gil?
Copied from original issue: mcveanlab/treeseq-inference#3