jwbelmon.bsky.social
#MD, #PhD | #Pediatrics | #Medical #Genetics | #Human #Genetics | #Cardiovascular | #Congenital | #Genome #Sequencing | #HealthEcon | #Causal #Inference
belmontgenetics.org
772 posts
1,658 followers
2,383 following
Prolific Poster
comment in response to
post
Can you tell what the three Institutes would be?
comment in response to
post
To be clear: Even greater damage will be done by the loss of federal government productivity. The workers who are losing their jobs were worth more than they were being paid! We are all poorer when roads, planes, and food are unsafe, when parks are closed, etc.
comment in response to
post
Been feeding Chat, Gemini, and Claude a stats problem from a textbook. They are all quite helpful in organizing the notation but all three made errors that would lead to completely wrong action if I were to naively use them on a real world problem. There has to be a human.
comment in response to
post
Claude didn’t get the algebra right either. It tried to factor out T instead of O when considering why the odds ratio is unbiased in a case control study. #Causal
comment in response to
post
a simple risk difference is potentially biased. Call the selection process O.
Both Gemini and Claude inferred the DAG as T->Y->O. That’s a chain not a collider.
The bias process is better represented as T->Y<-O. The idea is that {Y(1),Y(0)} is influenced by both T and O.
#causal
comment in response to
post
I ran the same problem by Claude 3.5 and it made the same initial mistake as Gemini.
The problem is the collider bias in a typical case control study. There is an exposure T and a disease Y. Selection of most cases Y(1) vs only a smaller fraction of controls Y(0) means that …
comment in response to
post
The conditional probabilities are not hard but the notation is hard to read (for me; lots of braces). DAGs are easy. I wonder what it’s trained on?
comment in response to
post
Very scientific
comment in response to
post
"The walkout was organized in part by an eighth-grade student who is not yet 14 years old, she and her mother said in a joint phone interview with The Washington Post."
www.washingtonpost.com/national-sec...