search

ben18785 / Selection_simulations

Simulating Wright-Fisher, Moran and Yule processes.
1 stars 1 forks source link

Other test results #7

Closed Armand1 closed 5 years ago

Armand1 commented 5 years ago

When we fit log series, lognormal and power law distributions to the WF simulations with beta ranging between 0 and -1 (NFD), and choose the best fitting distribution by AIC, we find that at relatively weak selection (beta < -0.5), the best fit is log series, but at stronger levels of selection that it becomes lognormal. This is great: exactly what we expect. It makes our discussion of this test of neutrality more reasonable: we can reject neutrality using this test --- but not at weaker levels of selection. Which is fine.

WF_SADs.pdf

Here the best fitting distribution is shown (yellow is log series; red is log normal)

ben18785 commented 5 years ago

Great. Selection with beta ~ 0.5 is still reasonably strong though (looking at time series plots), in that it’s easy to discern by eye. In the real series that we have, the effect of selection is generally less obvious; again diluting the worth of power law tests...

On 26 Dec 2018, at 13:38, Armand1 notifications@github.com wrote:

When we fit log series, lognormal and power law distributions to the WF simulations with S ranging between 0 and NFD -1, and choose the best fitting distribution by AIC, we find that at relatively weak selection (beta < -0.5, the best fit is log series, but at stronger levels of selection that it becomes lognormal. This is great: exactly what we expect. It makes our discussion of this test of neutrality more reasonable: we can reject neutrality using this test --- but not at weaker levels of selection. Which is fine.

WF_SADs.pdf

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.