I'm now applying a weight in the initial location averaging step that depends on the length of the bottleneck of each child lineage. Lineages with shorter bottlenecks are given more weight than those with longer ones because the more time that elapses during a bottleneck, the greater the chance of having migrated further away from the ancestor's origin. A next step would be to give more weight to more geographically consistent lineages, because we believe they are more reliably computed.
In the refinement step, the intial average locations of each clade are refined taking into account the location of the clade's parent with respect to a polygon defined by the children. I noticed that for clades with many subclades or basal lineages this necessarily resulted in an outlier being selected, which was however at least close to the parent. Now I have changed the logic to disfavor outliers by computing a position closer to the centroid yet in the direction of the parent.
These posts are the opinion of Hunter Provyn, a haplogroup researcher in J-M241 and J-M102.
