by Tyler Cowen
on January 23, 2013 at 12:24 pm
1. Belief in the great stagnation is the new normal.
2. The brutality of English leisure gardeners.
3. Are too many people photographing their meals? And can you surreptitiously film inside Disney World?
4. Is the (published) claim that most published results are wrong itself…wrong?
5. Is the Zara guy the third richest man in the world? On average they open a new store every day.
6. Why Cameron’s referendum talk doesn’t make much sense.
Well, the guy got the name of the EM (Expectation-Maximization) Algorithm wrong, but it’s a good read nonetheless. Type I error rate control is a real problem, and while they’re a pretty good literature on in (particularly in the statistical genetics world), a lot of researchers don’t appear familiar with it. Or maybe they ignore it because that’s the way their incentives are structured.
RE: #4 – there are two key assumptions that are made in the paper, one of which is stated explicitly in the supplemental materials:
1) The probability of a p-value being reported in an abstract is independent of its value, given that it is less than 0.05. This means that if the p-value is barely less than 0.05, it is as likely to be reported in the abstract as a p-value of 0.00001, rather than having language such as “the results were statistically significant” in the former case.
2) The p-values under the null hypothesis are uniformly distributed. This assumption will be violated if a study is designed to test for many hypothesis, and there isn’t an appropriate correction for multiple testing (see http://xkcd.com/882/ or http://www.niss.org/sites/default/files/Young%20Karr%20Obs%20Study%20Problem.pdf for explanations of this).
If either of these assumptions are violated, then the estimated proportion of null hypothesis wrongly rejected will be low.
Vaguely, it seems p-values in abstracts are advertised as a kind of selling point. Such as “we saw results with a p-value of X” where X is supposed to impress you.
To check my understanding, as far as I can tell, the Jager/Leek (JL) argument essentially boils down to the following:
(1) If the majority of reported results were false positives, then the distribution of their p-values should be basically flat over the range from 0 to 0.05.
(2) Actually, the distribution isn’t very flat at all: most reported p-values are less than 0.001 (see charts on p.10 of the arxiv paper)
(3) Therefore, most reported results can’t be false positives.
In a context where you actually observe all the p-values, premise (1) is fine. But it isn’t OK here: conditional on being wrong, 0.001 p-values are actually much more likely to get reported?
Did they assume textbook researchers who plan an analysis ahead of time and then only do exactly that test in exactly that way?
Multiple analysis and multiple tests are a fact of research practice. And, given 2 equally decent analyses but only space to report 1, you are likely to report the one with p<.001 ahead of the one with p<.05.
As a general rule, you shouldn't use the number of zeroes in the p value as an estimate of the size of the effect, or take the p value literally.
(I have not yet read the paper)
Statistician Andrew Gelman has a long post on #4 on his blog, with response by one of the authors.
#2. Wow. I had a community garden plot in Alexandria and sometimes you got more “advice” than you cared for, but generally it was quite nice. Tomatoes and such were occasionally stolen, but I had heard it was generally thought to be local homeless rather than jealous neighbors. Honestly didn’t begrudge homeless people my veges, but got super &^%$# ticked off when birds would take one bite out of each tomato.
#1: First point was valid, but probably a consequence of the fourth point, second point showed a frighteningly bad understanding of trade economics (whose only real insight was a restatement of the first point), third point was true but irrelevant and the fourth point was valid. So of the “four trends,” only one has any causal property.
#6: Five years of uncertainty? Now the referendum genie is out of the bottle, as Nigel Farage crowed, try at least twenty, if Quebec’s recent history is any precedent. Someone once said that in a single generation the English people ceased to be Romans and became Italians. As that was clearly not embarrassing and humiliating enough, it’s only taken them one more to transform England into Europe’s answer to Quebec.
In the Canadian drama that will play in Britain for the next several years, David Cameron plays the part of Robert Bourassa, using sovereignty as a threat to get what he thinks his people want from Ottawa (played by Brussels). Nigel Farage plays the part of Jacques “money and the ethnic vote” Parizeau, playing the provincial English toff better than Monsieur ever managed to do.
I find this comment rather revealing. Britain is a nation, not a province…right? The Euroskeptics’ fears are precisely that Britain is losing its sovereignty as more and more powers are centralized in Brussels.
I found the article rather weak. It only considered the obvious benefit of the EU to Britain (close economic ties), without even mentioning the arguments against–European courts overriding British courts, an additional layer of onerous regulations and laws, an ever-more expensive bureaucracy, etc. Of course if Britain’s EU membership is only positive with no downside, than raising the possibility of leaving can only be bad. But the reality is more complicated.
I am also not very convinced by the primary argument: that this is creating economic uncertainty. Britain leaving the EU wouldn’t automatically mean cutting economic ties. This would be in nobody’s interest and seems highly unlikely. Despite all the arguments that Britain leaving would be a radical, suicidal step, non-members Norway and Switzerland seem to be doing just fine (and have very close economic ties to the rest of Europe).
Anyway, what uncertainty there is exists with or without this referendum. It is not at all clear that the status quo (Britain as a member but with special exceptions) can be maintained. Many voices in Europe see Britain as a block to further integration and would like to give Britain the choice of developing closer ties or leaving the union in the next treaty.
I am no fan of Cameron and agree that he was probably forced into this for political reasons. But that doesn’t make the idea itself bad.
#1, I guess I’m always looking for an out. If GDP is stalled, look for other happiness. Buddhist happiness has good cognitive science cred, and does not require high consumption. Perhaps economics is not being normative in the right direction, given a wealthy, but slow growth, industrial economy.
So, based on 1., is Tyler a wind power advocate as a way out of TGS?
EROEI for new oil/gas is 11:1 – 15:1 maybe. Wind power is at ~20:1 and should only rise as increased economies of scale, modest incremental improvements in efficiency, etc.
I’m a wind advocate, but comparing EROEI’s of wind vs oil is a ridiculous comparison. They function in completely different roles. Oil is used primarily for transportation and wind is used for intermittent electricity production.
Comparing natural gas to wind is a somewhat better comparison, but even there comparing them using EROEI numbers is pointless since the cost of the energy inputs vary widely.
EROEI is the dumbest concept ever invented, since all energy is not equal. One needs to examine the value of the energy invested versus the value of the energy returned.
This concept is sometimes referred to as “ROI”.
Prattling on about EROEI does serve one useful purpose: it’s the easiest way to spot peak-oil crackpots or other such conspiracy-minded nutbags.
“The compounding mistake was a belief that globalisation would make everyone richer.” Hmm.
Also, why would a massive decrease in “the difference between energy extracted and energy consumed in extraction” only affect the “West”?
And let’s start measuring inflation and employment more accurately ASAP then, as it seems like the low hanging fruit of these points.
5.- And on a balance sheet of 10 billion euros..
How easily it is forgotten all three main UK parties promised a referendum on extra powers to Brussels at the last election. Given the discussions about the integration of the EU that will require treaty change as it is, a referendum was almost certainly on the cards in the next Parliament anyway. France and Germany (and the Obama Administration too for that matter) perhaps would do well to consider the popular feeling and reflect that this somewhat aggressive approach carries the great risk of putting people’s backs up and backfiring, ensuring an anti-EU vote.
Of course, given the history of the EU dealing with inconvenient popular votes no wonder everyone is afraid.
wondering about Zara future on the US since they don’t make XXXL sizes, do they?
When my Facebook friends go a day or two without posting pictures of their food, I assume they’re starving to death….
I’m waiting for the movie version of #2 — “There Will Be Blood Oranges.”
#4 MR is cited!
I have a real hard time in article 1 with the claim that inflation would be 6 points higher under calculation methods used in the 80s. First off, some of the corrections are legitimate; we really do need SOME sort of hedonic adjustment. But more to the point, it’s just impossible that ten years of 6% inflation wouldn’t be more obvious. Is there anywhere we can see a serious, non paranoid treatment of this subject?
I realize the article is addressing Europe, so maybe things are different over there. Americans have an unreasonable fear of inflation, and a paranoid streak that makes it hard to have a sensible discussion about it. So while I would put a little more weight on the validity of a European view, that degree of statistical manipulation still seems hard to believe.
“The compounding mistake was a belief that globalisation would make everyone richer.”
If 20, 30 years ago there had been polls in the western world asking whether people would agree to earn a bit less and use that money to lift hundreds of millions of people out of poverty in the third world, they would probably have answer answered:
“yes, sure!”, happy in the knowledge such a scheme wouldn’t get of the ground anyway.
Well it did happen, yet nobody is blowing their trumpet.
Is Peter Singer missing part of the issue? What’s the difference between a kid drowning in a lake across the road and another one starving to death in Africa? The same difference that is between jumping in and saving a life, and writing angry blog posts titled ‘it’s the economy, stupid!’
The believers in the great stagnation are almost certainly wrong. They are really not much different from any other faction proclaiming “the end of” something. Unable to see how humanity could possibly innovate beyond it’s present state.
Having read the article, the fact that energy is getting more expensive to produce seems a flimsy hook to hand a theory about the end of growth on. Our current forms of energy prduction are getting more expensive, but that incentivizes development of new technologies. We’re going to find alternative fuels, whatever they are. They might not be wind or solar, or even fusion. But when energy costs get high enough, there will be a collossal shiftto something other than oil, and then there will be a gold rush, and then energy will be cheap again.
Comments on this entry are closed.
Previous post: edX course with Banerjee and Duflo
Next post: Career advice from Richard Thaler
Email Tyler Cowen
Follow Tyler on Twitter
Email Alex Tabarrok
Follow Alex on Twitter
Subscribe in a reader
Follow Us on Twitter
Marginal Revolution on Twitter Counter.com
Get smart with the Thesis WordPress Theme from DIYthemes.