Open babarlelephant opened 2 years ago
Hi @babarlelephant,
Thanks for getting in touch and trying to use the package; as I'm sure you've noticed, the documentation is a work in progress.
Yes, it makes sense that setting nu=0.5 would lead to a more recent date if the time series has a bunch of zeros at the start. A couple of options would be to either: cut the start of the time series off, or use a time varying observation probability to allow for changes in the observation system as the outbreak was recognised.
I haven't had a chance to look through your XML carefully, but it looks like you
are using years as your unit of time in which case the "removal" rate mu=0.146
is a bit low. Also, I'm a little confused about how you are handling the origin.
I think it should be originTime=@origin.t
.
Please let me know how you get along, I'll leave this issue open in the meantime.
Cheers, Alex
Hi,
I'm a bit lost on the parameters, particularly mu and nu that appear to be critical.
The clockrate is fixed to 0.0008/subs/site/year
155 sequences from December 2019 and January 2020 and the epidemic curve from https://github.com/chaolongwang/SAPHIRE/blob/master/data/Covid19CasesWH.csv
The goal is to see how it affects the tMRCA
My xml is there https://github.com/babarlelephant/trash/blob/main/consensus-155.xml (the one with the sequences is in the same repo)
When letting the uniform prior [0,1] on nu the tMRCA converges to roughly the origin time (late October) but my epidemic curve has 0 cases until December 8.
When setting lower=0.5 for nu it converges to a much more recent date.
There is a short description of the parameters in https://github.com/aezarebski/timtam2/blob/main/src/timtam/TimTam.java
Thanks a lot!