The migration parameters estimated by IM (m1 and m2), which represented the number of migrants per mutation (m = m/μ), were converted to population migration rates (M = 2 Nm = θ m/2).
However, the migration parameter was estimated to be 0.0417 (4Nm3→2) from the pooled tobacco populations to the Eastern NC potato population.
Not all methods strictly take one of these two routes, but we mention this broad dichotomy here because it gives a good indication of the underlying assumptions, the amount of data required, the nature of the dispersal (or migration) parameter to be estimated, and the spatial and temporal scales over which each method is pertinent and its estimates reliable.
Migration parameters were estimated across all loci in terms of mutation-scaled immigration rates M (m/μ) and mutation-scaled effective population sizes Θ (4 Neμ) were calculated for each population, as implemented in Migrate-n v3.6 [ 71].
For Model 2, parameter estimates for migration rates and effective population sizes (based on θ) varied by population (Table 6).
This least-squares estimation of τ is reliable but was found to be unreliable for estimating the migration parameter, M [ 63].
Although the distribution for the RpL17A locus does not reach the zero value on its right tail, this gene shows the highest estimate for the migration parameter in the direction of L. whitmani.
The comparison of both parameter estimates made evident that migration rates obtained with IMa are lower than those inferred with Migrate-n.
Model M1 ranks higher than M2 for the triplets GPC, HSE, and JOG, but given that model M1 is an approximation of divergence with gene flow that does not explicitly estimate a migration rate parameter, we prefer to summarize parameter estimates from the M2 model.
Migration parameters have been estimated for all loci together for each population in different IM runs.
