How to choose ref_pop when running 'gimble makegrid' (-r)

[kuangzhuoran]

Hi,

I don’t understand the meaning of the -r parameter: "Population ID of reference population used for scaling",
as well as the difference between the three parameters: A, B, and A_B.

My command is:
"gimble makegrid -z Chr29.z -l makegrid -m IM_AB -b 64 -r A -u 3.03e-9 -A 10000,300000,6,log -B 10000,300000,6,log -C 10000,300000,6,log -T 10000,1180000,6,log -M 1e-7,1e-3,6,log -p 5 -f"

[GertjanBisschop]

Simply put, if you set -r A then gimble returns all other population sizes relative to the size of A. If you set -r A_B then all population sizes are scaled relative to the ancestral population size. Importantly this choice does not affect the result in absolute terms.

A, B, A_B are simply strings used to identify the populations in the sample range of models gimble works with: A_B refers to the ancestral population prior to a population split (divergence event). If there was no split (migration-only models) then only labels A and B are relevant.
From the documentation:

• A or B or A_B (for models DIV, IM_AB, IM_BA)
• A or B (for models MIG_AB, MIG_BA): here we only have migration from A to B or B to A, so there is no ancestral population

[KLohse]

A, B, A_B are the effective sizes (N_e) of populations A, B and their ancestral population (A_B) which gIMble co-estimates.
Since it is standard to scale the migration rate in units of N_e, i.e. M = 2 N_e m (the number of migrants per generation), we need to define a reference population to convert unscaled m into a scaled estimate. As Gertjan says, this choice is arbitrary (i.e. it does not affect the estimation only how M estimates are reported) and set by -r.

[kuangzhuoran]

Thanks for your reply!