I think these two tweets are saying the same thing (??)
-
-
Replying to @eigenrobot @DanFriedman81
Direction of replacement is different. The physician fixed effects are in the most complete model, and the physician race is left out because it's made redundant. He was saying the fixed effects I highlighted as important were left out.
1 reply 0 retweets 2 likes -
Replying to @RAVerBruggen @DanFriedman81
eigenrobot Retweeted Spotted Toad
I think obsessing about linear model-specific issues is a bad use of energy herehttps://twitter.com/toad_spotted/status/1296118945993416704?s=19 …
eigenrobot added,
Spotted Toad @toad_spottedOne way you know the doctor race study is bullshit is even very intensive interventions to influence birth outcomes (here, Obamacare paying for individual nurse home visitors for low-income pregnant women) have zero or negligible impacts on birth outcomes. https://www.acf.hhs.gov/sites/default/files/opre/mihope_strong_start_final_report_final508_3.pdf … pic.twitter.com/fUp78YHarRShow this thread1 reply 0 retweets 7 likes -
Replying to @eigenrobot @DanFriedman81
Very open to it being BS, but if so I'd really like to know what problem causes the result! I imagine it's a useful story with lessons for other studies, whatever it is.
1 reply 0 retweets 4 likes -
Replying to @RAVerBruggen @DanFriedman81
I think its failure is massively overdetermined unfortunately
2 replies 0 retweets 5 likes -
if black patients have higher underlying/latent risk and white docs are assigned to higher risk patients , unless you measure that risk very accurately, white doctor/black patient outcomes will be worse, even "controlling" for risk, bc drawing from a higher risk distribution
2 replies 0 retweets 11 likes -
The physician FEs should cover it if it's just white docs getting higher-risk patients in general, but if some docs see white lower-risk parents while also getting a more diverse set of higher-risk ones, that could create the interaction they found.
2 replies 0 retweets 3 likes -
Replying to @RAVerBruggen @toad_spotted and
To me the biggest hurdles for any theory are the physician FEs, the fact the effect is specifically on the interaction, and the fact you need a theory where healthier blacks are more likely to have black docs, which is somewhat counterintuitive.
2 replies 0 retweets 2 likes -
Replying to @RAVerBruggen @toad_spotted and
Is that counterintuitive? Wealthier people are more involved in physician selection, and it seems realistic for them to select doctors like themselves. Poorer people get whoever's around when they show up. (Whether this is empirically true is another matter).
1 reply 0 retweets 2 likes -
Replying to @knrd_z @toad_spotted and
Yeah, that's one of the most plausible explanations this endless branching thread has produced!
1 reply 0 retweets 3 likes
so at this point you've identified one reason the results are potentially borked and you havent even touched on model selection and specification, error estimation, or overfitting this is why I just dismiss this sort of thing out of hand
-
-
This Tweet is unavailable.
-
Replying to @DanFriedman81 @RAVerBruggen and
heh yes also data cleanliness
1 reply 0 retweets 4 likes - Show replies
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.