delt0r
commented
12 years ago Not sure i totally follow. If the model is not time invariant while selection is running, you cannot condition on end frequency. Basically the math just does not work.
However if what you want is to have a section of time where there is thing happening before the sweeps ends, this can be added.
But is something does happen far into the past. ie a bottle neck, the sweep may need to "originate" further back that this to meet the requirements of the conditions. This in this case the model fails.
More or less condition on final frequency, although the natural way to do things. Does not work out mathematically, in that its really computational difficult if even tractable at all.
At the moment, it seems that if we want to model a time-variant demographic scenario and selection together, we can only do so by conditioning on the starting time and frequency rather than the ending time and frequency. We were wondering if you could incorporate selection with demography such that we condition on the ending time and frequency, but in a simplified scenario, as follows:
We would like to model selection in a bottleneck scenario where the bottleneck has completed and the population now has an invariant population size so that we can condition on the end time and frequency after a bottleneck. For example, we have a simple scenario where the population size changes from 10^6 to 10^4 back to 10^6. In the second epoch of Ne=10^6, we would like selection to start and end after the bottleneck during the second Ne=10^6 period.