Today in "Tweets for no one"
It feels somehow wrong to me that Prais-Winston was published in 1954 while Newey-West was published in 1986...
It feels somehow wrong to me that Prais-Winston was published in 1954 while Newey-West was published in 1986...
Newey always feels like a bit of a lazy cop-out to me (Although I do still use it):
Yeah, we know OLS doesn't work real well with autocorrelation, but we'll just tweak the standard errors using some matrix algebra and everything will be fine.
Yeah, we know OLS doesn't work real well with autocorrelation, but we'll just tweak the standard errors using some matrix algebra and everything will be fine.
Whereas Prais "feels" more responsible:
Quasi-difference the whole damn model. Oh and here is a (sort of sophisticated) way to come up with an efficient AR(1) parameter so you don't have to drop an observation.
Quasi-difference the whole damn model. Oh and here is a (sort of sophisticated) way to come up with an efficient AR(1) parameter so you don't have to drop an observation.
If @Chris_Auld has any problems with this thread I want you all to know it's a big joke and I totally meant to misunderstand whatever it is I misunderstood.