Fig. 2

Ridge Analysis of the four-dimensional saddle surface representing the stationary point of a RSM model where the four selection coefficient parameter values minimize the misfit between Monte Carlo simulations and the data observed by Oliver et al. [33]. Here, the X-axis represents negative frequency dependent selection (T’s), the Y-axis represents additive selection (S’s) and Z-axis represents misfit for haplotype I (left) and haplotype II (right). The illustrated Ridge Analysis starts off with an arbitrary value of 0.75 for all four selection coefficients, and shows that the best fit with data occurs where there is no additive selection on mtDNA (S _I , = S _II  = 0) but where negative frequency dependent selection is equally strong on both haplotypes (T _I  ≈ T _II  > 0)