Construct simulation test set for paper

[halasadi]

(1) Show eems-ibd can infer parameters under the coalescent

m = 0.01, N = 10k, nsamp_per_deme = 50

(2) Infer barrier 10 generations ago

• IBD segments greater than 8 cM.

• Now eems

(3) Different Popsizes and Barrier 40 generations ago

• IBD segments greater than 4cM, m = 0.1, N =10k

You can see that a barrier is detected but the actual inference of parameters are way off. Thus, it appears that the data here is not informative of the actual demographic parameters but is informative of the relative differences.

Actual Popsize scalars: 0.65 0.39 0.03 0.22 2.2 3.6 0.7 8.2 0.23 0.07 7 0.32

Observed within IBD: 0.226122 0.275918 3.867760 0.642449 0.269388 0.138776 0.277551 0.306939 0.461224 0.866122 0.106122 0.232653

(4) Past Barrier. Suppose there was a barrier 40 generations ago but now has disappeared.

3.5-6cM

Not working.

8-Inf cM

Not working.

(5) New grid with new sample size

• 5x4 grid with nsamp_per_deme (haploids) = 24 (instead of 50 as before).
• Use LOD = 4 as cutoff instead of LOD = 1 (1 is very low, lots of false positives probably).