@sarahdoingthing AFAICT goodhart, berkson and explaining away are three distinct phenomena
-
-
Replying to @The_Lagrangian
@The_Lagrangian oh sorry I meant goodhart/Lucas as one relationship1 reply 0 retweets 0 likes -
Replying to @sarahdoingthing
@The_Lagrangian just unclear on how berkson & explaining away differ1 reply 0 retweets 0 likes -
Replying to @sarahdoingthing
@sarahdoingthing yeah actually the more I think about it the more I think they are the same thing2 replies 0 retweets 1 like -
Replying to @The_Lagrangian
@sarahdoingthing thinking out loud for a bit: I originally thought they were different because Berkson seems like it's a prob of subgrouping1 reply 0 retweets 1 like -
Replying to @The_Lagrangian
@sarahdoingthing whereas explaining away seems like general feature of probability: you need to conserve total probability1 reply 0 retweets 1 like -
Replying to @The_Lagrangian
@sarahdoingthing but actually explaining away applies specifically to the case where multiple hypotheses can explain one data point1 reply 0 retweets 1 like -
Replying to @The_Lagrangian
@sarahdoingthing so I guess subgroup membership can be explained by multiple things1 reply 0 retweets 1 like -
Replying to @The_Lagrangian
@sarahdoingthing yeah okay this is why Pearl calls it a "collider" problem1 reply 0 retweets 1 like -
Replying to @The_Lagrangian
@sarahdoingthing it happens in any case where you've got a causal graph like A -> B <- C. The arrows from A and B "collide" in C2 replies 0 retweets 1 like
@The_Lagrangian niiice
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.