# Thread: Iron Age Inspired G25 Calculators

1. ## Iron Age Inspired G25 Calculators

Like ph2ter, I have been working on an ancients calculator with G25 coordinates. These are inspired by the Iron Age references in the set, but the goal was to build a "world" analysis. My method was to use Past3 K-means clustering on the entire G25 reference setting the "K" size in the 20s (I forget my starting point). Some clusters had good Iron Age references that clustered with modern populations and for those I took the Iron Age references and formed a population of the type represented. Some clusters were defunct/subsumed (like the various hunter gatherer references clustered together) and I discarded those since they were represented in later Iron Age clusters. Some clusters were only "modern" because there are no ancient samples to represent them. For those, I used the moderns so they would be represented, given the goal of the calculator. Some of the clusters had Bronze Age or earlier representatives only, but still clustered with the moderns (showing some continuity from the era and forward). So, I kept those clusters with pre-Iron Age references. Finally, some of the Iron Age-only clusters were improved by adding some Bronze Age references to "complete" the cluster. I made two versions with the same references.

The first one is the "natural" one that results from this process, but European substructure doesn't come out strongly with it. Remember - this is trying to show the state of things from the viewpoint of the Iron Age and some of the current mixes weren't there, yet. I do like this one the best and should work OK, I think, with a mixed person or to see possible mixing sources. I may have "whiffed" on some names (apology ahead of time), but you can use the tags to right of the colon to see what was grouped in the cluster. I liked the results for this one using "pen=0" since it is intentionally looking for mixture and I tried to make the "lines" pretty distinct between the clusters. But, run it as you like. It is "unscaled".

The second one divides the "Europe" and "Mediterranean" clusters into some substructure that I got by applying the K-means clustering to just the first three dimensions of the PCA to those clusters. I think it works best with the distance penalty "on". It is "unscaled".

2. ## The Following 12 Users Say Thank You to randwulf For This Useful Post:

bmoney (11-20-2018),  Camulogène Rix (11-19-2018),  Capitalis (11-19-2018),  Hurricane (11-19-2018),  JMcB (11-19-2018),  lukaszM (11-19-2018),  michal3141 (11-19-2018),  Mykhailo (11-19-2018),  ph2ter (11-20-2018),  PoxVoldius (11-20-2018),  Robert1 (11-20-2018),  Username (11-20-2018)

#### Posting Permissions

• You may not post new threads
• You may not post replies
• You may not post attachments
• You may not edit your posts
•