Original estimate outside CI built with bootstrapping

[jaurbanChicago]

Hello Stephan,

I hope everything is well. How should I interpret if the original estimate for a split time between two populations (rCCR=0.5) falls outside the CI I built using the bootstrapping script from msmc_tools (20 replicates and using default parameters)? Image is attached.

Thanks in advance!

[stschiff]

Well, two things:

1. The CCR midpoint was never meant as a principled "estimate for the split time" in a statistically rigorous sense. It is simply what it is: the time-point at which our estimates of coalescence rates between the populations is half the estimates within the populations. For continuous splits (i.e. think of it like a drop in cross-migration rates or so), this is a meaningful time-point. For hard-splits, MSMC2 does not have the right model, strictly speaking. Take a look at the MSMC-IM paper for more discussion about this.

2. The time you're showing here, at least for humans and arguably many mammals, is very recent and likely at the boundary of what MSMC can measure. So that may contribute to the shift.