Thomas Lumley
@tslumley.bsky.social
Biostatistician. Baritone. He/Him.
Product of more than one country.
May contain nuts.
Product of more than one country.
May contain nuts.
Famous NZ comedian John Clarke had an extremely well-known farmer character called Fred Dagg,
(from en.wikipedia.org/wiki/Dag_(sl...
(from en.wikipedia.org/wiki/Dag_(sl...
Dag (slang) - Wikipedia
en.wikipedia.org
November 11, 2025 at 3:04 AM
Famous NZ comedian John Clarke had an extremely well-known farmer character called Fred Dagg,
(from en.wikipedia.org/wiki/Dag_(sl...
(from en.wikipedia.org/wiki/Dag_(sl...
I would suggest calling it Fred, but that's a very local joke
November 11, 2025 at 2:10 AM
I would suggest calling it Fred, but that's a very local joke
In my field, LLMs are quite useful because there are tasks that are slow to do by hand, easier to check, and not worth learning -- eg "give me the 18-25 populations from these StatsNZ spreadsheets".
The problem, generally, is the many tasks that satisfy the first two conditions but not the third.
The problem, generally, is the many tasks that satisfy the first two conditions but not the third.
November 9, 2025 at 7:17 PM
In my field, LLMs are quite useful because there are tasks that are slow to do by hand, easier to check, and not worth learning -- eg "give me the 18-25 populations from these StatsNZ spreadsheets".
The problem, generally, is the many tasks that satisfy the first two conditions but not the third.
The problem, generally, is the many tasks that satisfy the first two conditions but not the third.
Along these lines, Houston Rockets player Steven Adams has a famous sister with a bunch of Olympic shotput medals
November 9, 2025 at 6:04 PM
Along these lines, Houston Rockets player Steven Adams has a famous sister with a bunch of Olympic shotput medals
I think we've managed to do that to attach()
November 8, 2025 at 10:09 PM
I think we've managed to do that to attach()
Reposted by Thomas Lumley
“But Kartik, how do we know the genotypes are wrong?”
Trios! Both cohorts have ~1k parent-offspring trios that were recruited incidentally.
Applying genotype-level QC reduces Mendelian errors by ~60-80% (even in All of Us where they already did genotype-level QC on hom-refs!)
Trios! Both cohorts have ~1k parent-offspring trios that were recruited incidentally.
Applying genotype-level QC reduces Mendelian errors by ~60-80% (even in All of Us where they already did genotype-level QC on hom-refs!)
November 8, 2025 at 9:31 AM
“But Kartik, how do we know the genotypes are wrong?”
Trios! Both cohorts have ~1k parent-offspring trios that were recruited incidentally.
Applying genotype-level QC reduces Mendelian errors by ~60-80% (even in All of Us where they already did genotype-level QC on hom-refs!)
Trios! Both cohorts have ~1k parent-offspring trios that were recruited incidentally.
Applying genotype-level QC reduces Mendelian errors by ~60-80% (even in All of Us where they already did genotype-level QC on hom-refs!)
"Italian": fior di latte or lemon or artichoke?
November 8, 2025 at 1:31 AM
"Italian": fior di latte or lemon or artichoke?
Noether's got to be the *best known* woman mathematician from history, even so
(at least if you don't classify Ada Lovelace as a mathematician)
(at least if you don't classify Ada Lovelace as a mathematician)
November 7, 2025 at 8:44 PM
Noether's got to be the *best known* woman mathematician from history, even so
(at least if you don't classify Ada Lovelace as a mathematician)
(at least if you don't classify Ada Lovelace as a mathematician)
I'd do stir fry, but hash (cf colcannon) is another option
November 7, 2025 at 6:56 AM
I'd do stir fry, but hash (cf colcannon) is another option