If you are teaching graduate level Econometrics coming year, my new twin textbooks are now available on Kindle. Print copies are scheduled to be ready Aug 16 & 23. The Kindle version can be read on most devices, including phones, and PCs!
This morning, the Interactions Conference kicks off at UW in Madison, after a multi-year covid pause. We are excited to host a stellar group of both junior and senior scholars at the nexus of applied and methodological econometrics.
@jmwooldridge
Jeff, over the past two months I've had a continuing conversation with Steve Portnoy. He is written a follow-up paper exploring conditions under which unbiased must be linear. At first he thought it obvious, but his proof is advanced, and has conditions
@jmwooldridge
Three comments: (1) Report p-values, not "significant at the 5% level", as the p-value has more information. (2) Indeed, the 5% threshold has no scientific basis; it is a rule-of-thumb. (3) Researchers should use judgment when evaluating statistical tests, not just rote rules
@lihua_lei_stat
Lihua, you have no need to apologize, not for an unsubmitted draft, and not to an unhinged rant! I am amused that the exchange has produced a new term: "purported Twitter-proof"
@MarkAhrens
They are aimed at PhD courses. Methods rather than programming is emphasized. Code and illustrations in Stata, R, and Matlab are included in the text
@WhiteHouseCEA
Is there an economist in the house? Trend forecasting is dubious unless there is *structural* validity to the trend function. Indeed *within sample* a fitted trend can sometimes be a helpful visualization tool. Out-of-sample it is extremely dangerous.
@jmwooldridge
Jeff, I've tried to be consistent, to clearly distinguish errors from residuals. But I've noticed that in common discourse, it is very hard for most people. Slippage is so common.
@jmwooldridge
@CavaliereGiu
@lihua_lei_stat
Jeff, yes, the original version assumed independent observations. A referee thought it'd be nice to extend to correlated errors, which I did in the revision , and now regret.
@YujungHwang3
Legacy privilege is (unfortunately) not up for review in the Harvard/UNC court case, as it does not violate the law. Legacy admission completely skews Harvard admissions.
@jmwooldridge
Jeff, uniform (flat) priors are not uninformative. To see this, note that a flat prior on theta and h(theta) field different posteriors, if h() is nonlinear
@Mylovanov
This is wonderful news for the rule of law. I doubt he will get jail time. I don't know what other options are available, but I think it would be great if he was sentenced to a hefty fine plus community service.
@ArielKarlinsky
Ariel, thanks, I had not read the blog by Simonsohn, but I fully agree with it. Yes, it is not a new idea to recommend HC3/Jackknife, my paper simply adds to the theory
@jmwooldridge
@JohnMullahy
After seeing the new football coach arriving via a private jet, I am going to make the case that visiting seminar speakers be given the same treatment
@jmwooldridge
One way to generate an uninformative prior is to use Jeffries prior, which is proportional to the square root of p(1-p), if I remember correctly
@CavaliereGiu
@JEconometrics
Congratulations! Unfortunately, the formerly annual dinner has been downgraded to a pre-dinner reception. Giuseppe is a Jolly Good Fellow!
@adewed00
Yes, the book "Probability and Statistics for Economists" was written for a reader with no prior background in probability, statistics, or econometrics. The primary intended audience is first-semester PhD students in economics
@demasschuler
That is correct, by standard definitions. However, we could expand white noise to include all mean zero uncorrelated series, and then I believe mds is strictly nested