its probably the best they can do with the data available (and available methods in 2011 or whatever) i just dont think its enough
-
Show this thread
-
broke: structural models woke: credibility revolution bespoke: give up identification is generally impossible revel in your acknowledged ignorance and just tear down other people who think they know better with your familiarity with common pitfalls aka econometric socratism
2 replies 1 retweet 13 likesShow this thread -
Replying to @eigenrobot
is stats really not just a bunch of made up crap i only took the introductory class and had one of those profs who was all "don't worry about the math just memorize the formula and plug the right stuff in" which still makes me kinda angery to think about
1 reply 0 retweets 1 like -
Replying to @moon_of_a_moon
its real but most professors suck at it and even if your good its hard to do well
1 reply 0 retweets 1 like -
Replying to @eigenrobot
your professor was an awful teacher but the curriculum is awful too if youre not trying to be a specialist really hard to git gud without many years of work
1 reply 0 retweets 1 like -
Replying to @eigenrobot
sounds legit i mean it's kinda complicated and i assume its a "learn how then learn why" kinda setup but i hated it. didn't help that all the math profs i had at the time were awesome "here is how and why and alternate proofs of each" type teachers
1 reply 0 retweets 1 like -
Replying to @moon_of_a_moon
you can do those sorts of proofs in stats too but they usually dont for intro courses because they have to teach them to everyone not just numerate upper division math folks a second problem with stats is that proofs dont translate to practice very directly
1 reply 0 retweets 1 like -
Replying to @eigenrobot
the first part makes perfect sense, and the second part kinda does too since it seems more like a set of rules for applying math and the proofs for why they work probably have to assume a sort of idealized set of data for whatever they're proving or there's nothing to apply it to
1 reply 0 retweets 1 like -
Replying to @moon_of_a_moon
yeah that's about it. although ml is a major improvement in many cases at a certain cost
1 reply 0 retweets 1 like -
machine learning basically algorithms that make better predictions with fewer assumptions than traditional parametric models (you may have seen OLS/linear regression in your class, classic old style model) this at the cost of being slightly harder to interpret/prove theorems
-
-
Replying to @eigenrobot
ah gotcha that's what i thought wanted to be sure.
0 replies 0 retweets 1 likeThanks. Twitter will use this to make your timeline better. UndoUndo
-
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.