5 ms·

> For example, take the simple family tree above. If I’m person 28 on this family tree, and I’m making a genealogy, I would care about my own ancestors, but I m

by baq 3y ago

> For example, take the simple family tree above. If I’m person 28 on this family tree, and I’m making a genealogy, I would care about my own ancestors, but I might not care much about my ancestors’ siblings (my aunts and uncles) and their descendants. So most of the family tree won’t be present in a genealogy database Or, the data is correct and a lot of lines of people just die out without offspring.

yorwba 3y ago
Or most people can't trace their lineage all the way back to 1809 and some ancestors just spawn into existence (or descend parthenogenetically from a single parent) at a later date. Maybe those missing ancestors could be imputed by sampling a random other individual with known parents born in the same year.
- dhosek 3y ago
  I think that’s likely it. One should instead build the tree backwards starting with the most recent members of the database and then find their earliest known ancestors. These can then be assigned the magic trait based on the likelihood of it being present based on their birth year cohort (or more reasonably birth decade or some other larger bucket).
gwern 3y ago
No, that can't explain his result showing the witch gene going extinct despite, by stipulation, zero fitness impact on individuals but 100% transmission. Because the human population did grow over that time, so every deadend must have been balanced out by even more children by the non-deadends. And since he starts at hefty fractions like '1% of the entire population' (and goes up to 20%), we should be well beyond any issues of very small number effects or gambler's ruin. There has to be something wrong with either the data or code. (It almost looks like a sign error!)
sct202 3y ago
The linkage file is probably missing a lot of linkages. As someone who has done some genealogy research, at a certain point you just start guessing based on last name, age, region and hope its right. There's also a big trend of people getting to a point where there are multiple options (similar names, ages, region) and picking the most well documented lineage to hook their genealogy on to.

by baq 3y ago

yorwba 3y ago

Or most people can't trace their lineage all the way back to 1809 and some ancestors just spawn into existence (or descend parthenogenetically from a single parent) at a later date. Maybe those missing ancestors could be imputed by sampling a random other individual with known parents born in the same year.

dhosek 3y ago
I think that’s likely it. One should instead build the tree backwards starting with the most recent members of the database and then find their earliest known ancestors. These can then be assigned the magic trait based on the likelihood of it being present based on their birth year cohort (or more reasonably birth decade or some other larger bucket).

gwern 3y ago

No, that can't explain his result showing the witch gene going extinct despite, by stipulation, zero fitness impact on individuals but 100% transmission. Because the human population did grow over that time, so every deadend must have been balanced out by even more children by the non-deadends. And since he starts at hefty fractions like '1% of the entire population' (and goes up to 20%), we should be well beyond any issues of very small number effects or gambler's ruin. There has to be something wrong with either the data or code. (It almost looks like a sign error!)

sct202 3y ago

The linkage file is probably missing a lot of linkages. As someone who has done some genealogy research, at a certain point you just start guessing based on last name, age, region and hope its right. There's also a big trend of people getting to a point where there are multiple options (similar names, ages, region) and picking the most well documented lineage to hook their genealogy on to.