French

• Branches of the era. We take the ancient samples whose date falls inside the era, and compare the lines as they stood 1,500 years before the era ends.
• A culture’s profile is the share of its lines on those branches. The number of samples does not change a weight: a culture of seven men weighs as much as one of a hundred and twenty.
• A shallow test. An ancient man typed above the cut is spread over the branches under his node, by how many men each holds now. Today’s kits are treated the same way.
• Who is admitted. A culture takes part when its sharply typed men cover 25% of the population on their own: branch by branch we add the smaller of the two shares, the culture’s and the population’s. One dug within 550 km needs only 0.5%.
• The shares. We pick the weights of the cultures so that their mixture comes as close as it can to today’s set of branches.
• Look-alike cultures. When cultures share almost all their branches, the data cannot say whose kin the men are. So we do not split the share between them: they get one bar.
• The range. We recount everything on random subsets of the samples. In brackets are the bounds the share keeps in nine runs out of ten. A share that can fall below 5%, or below half of itself, is not shown at all.
• “Fragile”: the share rests on a few samples, or a random shuffle of cultures gives as much.

What it does not mean. This is kinship of male lines, not the origin of a people. The female lines say nothing here. Who takes a test is a matter of chance, so the shares compare between countries, not with a census.