Statistical Modeling, Causal Inference, and Social Science


I had a dream last night that I was at a conference, then I was going to the basement of the convention constructing to the grocery store, then I was at my mother’s house and she or he told me I needed more additives for dinner, so I went back to the basement, but the stairs didn’t go all of the way down, so I had to kind of shimmy down the wall, then when I was down there with a supermarket cart I overheard some old guy, a grocery store employee, speaking about how much he hated his boss and he was only there because his boss was paying him extra for overtime cos it was Christmas, then the old guy and an old woman were having a big scene where maybe they were gonna get in combination again and that they finally commit to do so though they comprehend it could never really exercise routine, then to get down from that wall I’m balancing atop a ladder that’s gradually establishing up, then I’m outside in a plaza and that man and girl are doing a similar thing but they’re acting in some type of public play and lots of people are looking at, and then other actors pop out and everybody’s applauding. I remember thinking in regards to the play that the actors were fine—I wasn’t sure in the event that they were doing it at no cost or as a part of a paid performance—but that something that wasn’t so convincing was that they were in their 60s or 70s however the woman had a bit baby. But then it turned out that she was really a young actress, some celeb, and she or he removed her old lady makeup to the acclaim of the group. Then I’m jogging away from the plaza down the street—here’s some unnamed U. S.

city—and then I’m back in that constructing I’m going up stairs or maybe I get off the elevator and there’s some shelf, form of an indoor ledge that’s really vital, I think it has some form of keyboard and screen with interesting guidance, but which you could’t stand on the ledge, you could only hang on to it by your fingertips, and as I’m seeking to do so—I must be cautious because if I slip there’s a limiteless drop below and I’ll just die, so I’m form of hooking my leg around the wall on the right side of the platform so I don’t fall—a Japanese guy comes up the steps, easily grabs onto the platform and is maintaining on together with his fingertips and as an aside says to me, Andrew Golemann—or whatever thing like that, I don’t bear in mind exactly, but I do bear in mind it was some misspelling and mispronunciation of my name—they’ve your name wrong on all but one place on wikipedia, and I replied, no, my name is really Gelman and he said he’d go fix that page and I asked him to support my back in order that I didn’t fall off the ledge, and I guess I didn’t fall off, because a higher thing I be aware is that I was shopping at some kind of online conference schedule, after which I was speaking on the telephone to Jeff Lax saying that I was shocked not to see him at this convention I knew it was a political technological know-how conference because it had events similar to “8:20 9:20 International Relations Book Sale” and I was telling Jeff how I’d been in California and visited the University of California and seen Jas, and Jeff said, Jas?, and I said, Jasjeet Sekhon, and Jeff said but he’s in Seattle and I said yeah but he was visiting California that day and we had lunch or even it was dinner and Jeff said where did you have got dinner which seemed kinda weird like why would you care and anyway Jeff was saying he couldn’t go because it conflicts along with his teaching after which I looked carefully at the agenda, it was only one day long and anyway I had to educate Tuesday morning that is sensible, the conf was on Monday because all that other stuff had happened over the weekend in the dream; there have been a flight along the style and I said I hope he had a flight and Jeff said the flight may be dear and maybe the college travel agent could help—really I don’t think there’s a school travel agent—so I said I’d call the airline and check, because that resolved things last time I was uncertain if I had a flight already scheduled, and around then I awoke, I’m not sure why as it was still 20 minutes before my alarm was scheduled to go offYou can draw lots of conclusions from the above dream adding that not anything in it was very scary other than that bit about needing to hold by my fingertips, but what I noticed upon waking up was that during my dream I’d been at a convention!I don’t go to meetings anymore. Yesterday I spoke in France but it was remote, I wasn’t truly in France and indeed I was told that many folks in the viewers didn’t even speak French. There was no coronavirus in my dream. And then I found out that I’ve not yet had any dreams with coronavirus in them, or at least I don’t remember. I was going to say that this increased uncertainty comes from the Fivethirtyeight model assuming that 2020 is a particularly unpredictable election, but I guess not, as a result of here’s Nate saying, “this year’s uncertainty is about average, that means that the old accuracy of polls in past campaigns is a fairly good guide to how correct they are this year. ” So now I’m really undecided where that enormous uncertainty in the Florida prediction is coming from.

See also  The Official Transformers Collectors' Club

Maybe their between state correlation in forecast errors is too low, in order that they need artificially high uncertainties in the particular person states to get the correct countrywide error?Looking at specifics has worked for me before. After the 2008 election I made some maps posted on the Fivethirtyeight site, definitely using poll data to estimate how alternative people would vote. The next day, I got slammed by blogger and political activist Kos for some implausible estimates from my maps. And Kos was right!It was good for me that I plotted a gaggle of maps—not only, say, some regression coefficients—as a result of that gave enough information for an intruder to catch complications with what I was doing. After a pair months of retooling, I posted some more advantageous maps based on a better model. Since then we’ve done more; see for instance here.

One problem when decoding these forecasts is they don’t constitute all feasible results. Everything we’re doing is in terms of the proportion of the two party vote, so, if there’s any severe third party problem, all bets are off—or, maybe I should say, our bets are off. Also, both forecasts are framed as Biden vs. Trump, so if either candidate dies or is incapacitated or is otherwise removed from the ballot before the election, it’s not quite clear how to interpret the models’ chances. I mean, sure, we could just take them to constitute Dem vs. Rep, and that probably wouldn’t be so remote, but this opportunity does not look like protected in either model.

And what if the votes never get counted as a result of a mega version of the 2000 Florida dispute over vote counting?Again, it’s not clear how this would be dealt with in a hypothetical bet, but I guess that some adjustment would be essential. Elliott Morris estimates that we should reduce Biden’s win likelihood that’s win likelihood, not vote share by about 7% if we want to encompass higher than usual ballot rejection rates in the model. I’d say that the present models do implicitly account for the opportunity of small disputes equivalent to those 30,000 votes that were never counted for Gore in Florida but nothing huge corresponding to 20% of the ballots in the state getting lost or ruled invalid. On Twitter a person regarded our results as pessimistic, as we point out misspecified models and in real life we can assume that none of the models is the actual data producing mechanism. With misspecified model we mean opposite of well specified model that doesn’t are looking to the true data producing mechanism, and of course the quantity of misspecification matters.

The discussion about well specified and misspecified models holds for any modeling approach and it’s not unique for cross validation. Bengio and Grandvalet had used just the term outlier, but we wanted to emphasise that outlier is not necessary a belongings of the information generating mechanism, but more of anything it really is not well modeled with a given model. Here’s a presentation, Exaggerated Claims Undermine Science by Ignoring the Scientific Method, by Rob Kass, a statistician who through the years has done a lot of interesting work on statistical theory and applications, especially in neuroscience. A few years ago, we discussed Kass’s thoughts on statistical pragmatism. And here’s a dialogue of a few papers by Kass and Brad Efron, which may be the first time I quoted Hal Stern that “the large divide in records is not between Bayesians and non Bayesians but rather between modelers and non modelers.

” That’s similar to the other Hal Stern quote that “what’s vital in a statistical method is not what it does with the knowledge but what data it uses. ” I guess Hal has a undeniable style together with his aphorisms. In his talk, Kass goes into details on a social neuroscience instance, discussing among other things the often misunderstood distinction among correlation and causation. As an individual who’s done numerous work in survey analysis, I’d also like to emphasize that correlation does not even imply correlation. Also, a minor thing: Kass discusses how a regression “controls” for variables.

I prefer to say “adjust for. ” But that’s all minor. Overall I recommend Rob’s talk in that it connects ordinary issues of technology to more precise questions on what’s information. This study explores the impact of women’s access to reproductive healthcare on labor market alternatives in the US. Previous analysis finds that access to the contraception pill not on time age at first birth and greater access to a school degree, labor force participation, and wages for women.

This study examines how access to contraceptives and abortions influences job mobility. If women cannot handle family making plans or doing so is heavily based on staying in one job, it is more difficult to devise for and take risks in their careers. Using data from the Current Population Survey’s Outgoing Rotation Group, this study finds that Targeted Restrictions on Abortion Providers TRAP laws increased “job lock. ” Women in states with TRAP laws are less prone to move between occupations and into higher paying occupations. Moreover, public investment for medically necessary abortions increases full time occupational mobility, and contraceptive coverage coverage increases transitions into paid employment. The prestigious multidisciplinary MIT Media Lab built a “personal food laptop” that worked so poorly that demos needed to be faked Theranos style .

. . According to Business Insider, the assignment—a plastic hydroponic grow box full of “advanced sensors and LED lights” that might supposedly make it feasible to reflect crop circumstances from any part of the worldwide—was a sham, with MIT’s Open Agriculture Initiative director Caleb Harper resorting to faking demos . . .

See also  How to Become a Successful Young Entrepreneur with Pictures

According to Business Insider, Harper directed an email soliciting for remark to an MIT spokesperson, who “didn’t provide a comment. ”Some leaders of the tutorial establishment have said that people who do original studies deserve more credit than mere critics, as an long-established study requires creativity and may advance technology, whereas a grievance is at best a footnote on present work. But I disagree with that stance. Or, I should say, it depends upon the customary study and it is dependent upon the grievance. Some common reports do advance technological know-how, while others are empty cargo cult workouts that at best waste people’s time and at worst can send entire subfields into blind alleys, as well as burning up millions of dollars and promoting a kind of quick fix Ted talk considering that can distract from real efforts to solve important complications. From any other course, some critical work is inconsiderate formal replication that sidesteps the scientific questions handy, but others—reminiscent of those of Nelson and Simmons linked above—are deeply engaged.

Sometimes. Not always. And often it’s controversial. For example, is Alexey Guzey’s grievance of Why We Sleep more effective than Matthew Walker’s book?I don’t know. I really don’t.

Yes, Walker makes errors and misrepresents data, and Guzey is contributing a lot by monitoring down these particulars. Any future researcher wanting to follow up on Walker’s work should certainly read Guzey before going on, simply to get a feeling of the evidence really is. On the other hand, Walker put in combination lots of things in one place, and, though his book is fatally flawed, it arguably remains to be making a crucial contribution. Sleep—unlike beauty and sex ratio, ovulation and voting, embodied cognition, himmicanes, etc. —is a crucial topic, and even though Why We Sleep misfires on many events, it can be making a real contribution to our knowing.

Look at it this fashion. Sometimes—repeatedly—researchers go into a assignment strongly believing that their major speculation is correct. In that case, fine, do a small among person study and it’s most unlikely that the results will basically contradict your speculation. In that case, the error in the usual paper is subtle, it’s the claim of sturdy proof when there’s no strong evidence. Then when the replication finds no strong evidence, the researchers remain where they started, believing in their usual hypothesis. It’s hard for them to pinpoint what they did wrong, as a result of they haven’t been considering concerning the distinction between proof and truth.

From their standpoint, they’ve broken some arbitrary rule—they’ve “p hacked,” which is set as silly as any other arbitrary rule of “p under 0. 05” that that they had to follow in advance. They see methodologists as like cops or as our nudgelords would say, Stasi and they care less about silly statistical rules and more about real science. 3. The Fabrigar et al. paper seems fine in a common sense, but I don’t think they struggle enough with the idea that outcomes and comparisons are much smaller and less steady than traditionally imagined by many social researchers.

To increase some old examples, it’s a mistake to come into the evaluation of an experiment with the expectancy that women are 20 percentage points more likely to vote for Barack Obama during a undeniable time of the month, or that a small intervention on four year olds will augment later adult income by 40%. Statistics based science is quantitative. Effect sizes matter. However, it happened to me that many statistician might like to be more aware of opportunities that such world research groups might offer. Most I think work at a single establishment where the analyses they get to be concerned in have a single data set at least at anyone time limit, a single analysis group is concerned in the task and with the unlucky force to get whatever published in a journal.

Eventually the research encounters the usual lower than adequate peer review from journal editors and reviewers. Then if post peer review occurs, some concerned in the analysis “demand” the wagons to be turned around and all individuals remain inside. Recently, in a series of talks by members from OHDSI at the digital JSM2020, some real modifications in alternatives to the above for statisticians seemed obvious. Briefly, in place of a single data set there are more than one sources of knowledge sets, summaries of the separate effects are made accessible for contrast and comparison, the researchers are often from a couple of associations around the world, there is a methodological group that can be drawn on for true expertise and code on github and peer review can doubtlessly be a part of the technique by others in OHDSI not at once involved. Still there seems that unlucky force to get whatever thing posted in a journal, at least for plenty in OHDSI. Someone asked me the other day whether a corporation could run for president.

I said no. The closest to that would be The Space Merchants. And that that made me think . . .

where are the collaborative novels?I’m not talking about ghostwriters, or about that book by James Patterson and Bill Clinton which I assume had a third collaborator who really did the writing. I’m talking about actual collaborations. Other than The Space Merchants and a pair other books by those two authors, the one example I can bring to mind offhand is The Gilded Age, which I’ve never read cos it’s not supposed to be so great.