That was stalking, but was it really an attempt at an attack? The woman moved first, probably startled the tiger.
I was sure the second link would be the person calling 911 mistaking a raccoon for a tiger.
Animal Identification Guide
Raccoon: black mask, striped tail
Tiger: N/A because by the time you see it, it's already too late
The Gahanna Lion
https://usatoday30.usatoday.com/news/nation/2004-05-04-lion-usat_x.htm
What they should do is release a tiger inside Apple headquarters and see what happens. For science.
Simultaneously. It has to be, otherwise it's not totally valid.
You'll need an extra tiger, but no price is too high. For science.
Right now Pence is holed up in a hotel room in Lima and the President is about to go on TV to say something about Syria.
I could have sworn that last week he said he was pulling us out of Syria.
It's like hiring Michael Cohen to pay off a woman. It works until it doesn't.
14: The speech Trump given actually makes some kind of sense if it's actually followed through consistently. Which maybe GB and France can do on their own when Trump flakes.
Though presumably they won't. And the Syrians are saying their (Russian-made) air defenses shot down a bunch of Tomahawks, which will be a landmark if true.
And some baby birds are crying unceasingly outside my window.
OT: on the tube this morning... "We apologise for the long wait this morning but due to staff shortages we are running less trains today. I'm sorry, _fewer_ trains."
Les train, les train.
Trump just tweeted the phrase "Mission Accomplished," so I guess we can all rest easy now.
23: Or maybe it's a little too obvious.
I think trolling the Deep State, not us.
Not that I favor escalation here, but it strikes me as pretty half-assed to blow up some warehouses outside of Damascus while leaving the Presidential Palace unscathed. If you are going to attack Assad, maybe you should, I don't know, attack Assad? At least it would be a coherent strategy.
The way attacking Saddam and Qaddafi was a coherent strategy. Decapitating Syria would just make an impossible clusterfuck more fucked.
Whereas limited strikes* in response to particular uses of chemical weapons might actually have the effect of restoring the norm against chemical weapons use, which is well worth doing.
*If consistent and inflicting significant pain, neither of which has so far been the case.
I'm not even saying kill Assad. If you bombed the palace he'd be holed up in a bunker, I'm sure.
But you could wreck his office, mess up some command and control facilities, destroy symbols of his power, and make him feel genuinely concerned with his safety. I think that would go a lot farther toward "inflicting significant pain" than whatever it is we did.
29: Half of Damascus has been out of regime control for years. Assad clearly has grit. Given overthrowing Assad clearly isn't actually an objective anyway, I think strikes should do such damage to military targets as clearly to outweigh any gain from using chemical weapons. AIUI Assad's most critical lack is actually personnel.
If there are good reasons to do a thing like this, it's incumbent on the Exec to send someone over to Congress to make the arguments. They could explain that we don't actually want to harm Assad or his warmaking capability, lest we upset his major ally, but just want to have an expensive fireworks show. Again. That'll get 60 votes in the Senate, right?
Its easy to have grit when you are sitting in a palace. Take away his palace!
To be clear, everything I'm describing is hypothetical. I fully expect this to be nothing but a fireworks show.
Right. I just want to change the venue on the show.
Blow up his palace and you're literally just feeding him grit with his morning yogurt.
To be more precise, I think these types of police action bombings would be more effective if they were used to make it clear that heads of state are to be held responsible for chemical weapons violations.
Decapitating Syria would just make an impossible clusterfuck more fucked.
What about having a secret affair with Assad impregnating him, and then getting Michael Cohen to give him $1.6 million dollars.
Maybe.
This situation is theoretically interesting to me as a very rare case where I think America's airpower-based Middle East policy might in principle be capable of achieving something.
Although actually I think Assad would cost more than $1.6m.
That's just for silence. If he kept the baby, he could still sue for child support.
And he'd sue for more if, let's say, his house were to get blown up.
I think he'd be happy if he had a baby with an actual chin.
THAT KIND OF THINKING LEADS DOWN A DANGEROUS ROAD.
OT: Taking candy from a baby feels horrible. Swiping Skittles from a tween does not.
Anyway, the Syrian airstrike made Alex Jones cry.
The best accomplishment of the Trump presidency so far.
45. I don't follow Alex Jones and have no idea why he is upset to the point of crying about the Syria air strikes. Does anyone out there know? Something about how the previous ones were faked to get us into war with Syria? O RLY?
All the sites that do "war stuff" are crammed with Russian trolls in the comments today.
47: I looked it up very briefly, and yeah, Jones seems to be concern-trolling that Trump has been misled into bombing by a False Flag Operation (presumably thanks to the Deep State? these things are hard to follow).
The Deep State woke me up five minutes before Stanley.
If that's really even you, Mobers. Could be a decoy.
British intelligence can prove I'm me.
Deep State
The most dangerous safety school.
||
Big event here tomorrow, and it's been a long and trying road to get here and I'm totally pooped.
|>
Did you remember to bring the frankincense?
Oh man, I thought I was supposed to bring the myrrh?
You're either 8 months early or 4 months late. Either way, I wouldn't stress.
I'm to give tours to ex-presidents and such.
||
I assume everyone has already seen this? Because it really is amazing.
||
To discourage generals from taking such reckless actions, Song sent them a signal by according an unconventional treatment to Hao Lingquan, a general from the Anxi protectorate who had managed to obtain the head of the Turkic leader Mochuo, killed by Bayirku tribesmen in 715. Presenting the head to the court, Hao believed that he had performed a great service to the emperor, thus deserving a generous reward. But Song Jing bitterly disappointed him. Song not only drastically reduced Hao's reward, but also decided to delay his promotion to Commandant of the Right Military Guard for more than a year. An enraged Hao staged a hunger strike to protest and died soon afterward.|>
Severed heads are like hiring prostitutes to pee on a bed. Neither should be a gift unless asked for in very specific terms.
Give me the peed upon head of Alfredo Garcia.
Bring me. It's been a while.
Also I will fight to death you motherfucking stupid ass copy editors who automatically reject Oxford commas. Fuck you, you ignorant ass motherfuckers. You didn't even consider how it changed the fucking meaning. I'm going to have to pass by that fucking label I wrote that you fucked up every fucking day I hope you die of pox.
Maybe you should hope they die of something that isn't contagious?
You're probably right.Sorry, I've been drinking. What would you suggest?
The specifics don't really matter. The important point is, only plan biological warfare when sober.
69: Drinking is usually what I suggest.
And biowar is Barry's established preference here.
64: his failure to obtain the Right Guard was no doubt due to his being in bad odour at Court.
I'm going to assume ajay was struck by lightening right after hitting "Post" on that.
78: He had lost the mandate of heavyen.
79 is outstandingly virtuous.
This is quite possibly the best one of these.
OT: I am deeply annoyed by this article and sharing my annoyance.
He starts by arguing, correctly, that Pearson correlation has weaknesses and is commonly used in cases where it should not be used. He then argues, again correctly, that his preferred statistic can do better than Pearson in some circumstances. But he seems to think that nobody has noticed this before and that everybody was just plugging away computing Pearson's r while waiting for a physicist to come develop a better way. Maybe there's some work done on this problem after 1860 and before somebody decided Eugene Volokh was somehow representative of current thinking in statistical analysis?
83: Physicist being arrogant? Say it ain't so!
83. That seems overly negative. The guy is a statistician, his (now deceased) wife was an astrophysicist. They used distance correlation, which he correctly suggested was proposed as a useful alternative correlation method in 2007 (he cites a paper), to extract relationships between galaxies. He (they?) also used it to debunk a simplistic article by Eugene Volokh on murders vs. gun ownership that was published a few years ago.
Moby, did Donald Richards kick your dog once or something?
I don't have a dog. He failed to kick the dog that bit my ass.
I'm negative because he's ignoring that anybody who knows what they are doing will test to see if the assumptions for any technique are reasonable, because he doesn't mention any of the many other ways of testing for a relationship that improve upon Pearson's r, and because he talks about the distance correlation as a panacea which is only very slightly less stupid than treating Pearson's r as one.
I'm even more annoyed now that I see he's a statistician. If dropping into a new field and saying "I am so right, you are so wrong" is the characteristic failing of physicists, using overly simplistic strawmen as a contrasting point when proposing an improved statistic is the characteristic failing of statisticians.
Will this make Moby even angrier???
https://www.motherjones.com/kevin-drum/2018/04/hooray-for-statistics/
"My new method is a huge improvement over doing things that have been known to be horribly wrong for over a generation, therefore I am right."
Maybe he just wants a cookie?
I think a better message would be that the Washington Post shouldn't publish statistical analysis done by a lawyer without review by somebody who knows what they are doing. I don't think that adding more tools to Excel would result in anything but different types of errors if you don't have somebody training or experienced involved.
That makes too much sense to be published in a newspaper.
92: The Full Employment for Statisticians Act.
I doubt it. I think it would result in fewer statistical claims being published. Which is helpful compared to having Eugene Volokh's conclusions being published as if they were science.
using overly simplistic strawmen as a contrasting point when proposing an improved statistic is the characteristic failing of statisticians.
Shouldn't they be especially skilled at arguing against robust null hypotheses?
It's a lot easier to argue against really shitty ones.
81 is good and I love the poster's twitter handle which makes it even better.
But now I'm having a little sad b/c I *just* came from the Kevin Drum post about distance correlation, which I did not previously know about and thought was very clever, and was enjoying a little "hey-neato-lookit-that" trip but here already I find that I am supposed to not be impressed, so ride over. Aw.
I'm not saying it wasn't a neat use of statistics.
||
Londoners, pinging about a potential meetup. I'll be in London from the 20th to the 24th; my wife and I are available for a meetup any of those nights, with the caveat that Friday I'll probably be pretty out of it. We had discussed having one Saturday night if that still works for you all. We're staying in the City but are open to anywhere we can get via Tube. Email address is in the link--I don't think I have any of yours.
|>
shouldn't publish statistical analysis done by a lawyer without review by somebody who knows what they are doing.
This feels like way too earnest a comment, but statistics need to be part of the law school curriculum. Not because they come up all that much (they don't) but because when they do come up they are either needlessly and pointlessly ignored, because math is scary, or used incorrectly to support a predetermined conclusion, because making sometimes maybe questionable but reasonable-sounding arguments to support predetermined conclusions is kinda what lawyers do. Either way there needs to be some baseline competence in the (existing, imperfect) science of causation in a field where people deal all the time with issues of great importance where the question is "did x cause y."
That law professors are instead, as a whole, way more into mumbling non-empirical nonsense about "incentives" (newer version: non-empirical nonsense about "market power" and "inequality") , or being tiger moms and publishing novels, is yet another reason why the law schools are a gigantic academic embarrassment.
I mean, if statistics had been part of the curriculum, my law school grades would have been worse. That would have sucked for me. But definitely good for society.
Isn't it better for the attorneys' souls that they don't even know enough to know when they are lying with statistics?
Moby is right that the idea of alternatives to the Pearson r is nothing new (some of the alternatives are literally 100 years old), but I wonder if it's the journalist who's at fault? They always oversell the novelty of everything.
I don't think that adding more tools to Excel would result in anything but different types of errors if you don't have somebody training or experienced involved.
At least we can agree on that! At the least clutter up Matlab or R with them. (IANA statistician, so for all I know they both have 47 different ways of doing correlation.)
I don't know about Matlab. R is cluttered with literally everything anybody ever thought of.
90. "My new method is a huge improvement over doing things that have been known to be horribly wrong for over a generation, therefore I am right."
"2007-2018 is not precisely a generation, Lord Copper."
110. s/R/every Linux app/
111.1: I don't know who Lord Copper is, but I was thinking of methods that were old when the 2007 paper came out.
R's built-in correlation function supports three: Pearson and two alternatives (Spearman rho, Kendall tau). You have to install a package for distance correlation. I realize that as I write this nobody cares, but it's less work to hit "Post" than to delete it.
Question for all the statistics nerds who are here now: I've been using this margins package to get average marginal effects of a logistic regression, but it always returns the warning "'weights' used in model estimation are currently ignored!", I assume because it's a multivariate regression. Is this method flawed-but-best-available, or irredeemably flawed? I'm trying to get a decent estimate of the real-world impacts of the significant independent variables, and need an alternative to regression coefficients.
(amusing extract from another part of the linked documentation: "(e.g., what if all women were instead men? what if all democracies were instead autocracies? what if all foreign cars were instead domestic?)")
I think they are referring to sampling weights, so that would only apply if you did that for your model. In SAS, I would use Constrast and Estimate statements to get the estimates you are talking about. I have no idea how to do that in R. With logistic regression, you don't want the coefficients anyway. You want the odds ratios.
If Buicks were Porsches, Motown would cry.
116: Oh, weights for each datapoint like in Census data? Yeah, I don't have that. Thanks.
I know I don't want the coefficients, but I've been told AME is a good alternative? I guess I need to look into odds ratios more though.
This is an admirably serendipitous thread.
I think it means sampling weights, and that they've just not gotten around to implementing it yet.
||
Some students are filling out paperwork to work with me this summer - ie really common paperwork for summer independent study - and
(1) it's not online
(2) To indicate the semester, the form helpfully says: 200_.
|>
Cross out all printed digits and give the Jewish year.
That was me -- I'm in a new office. In a building that was name-checked in a Bowie song.
The song is Diamond Dogs -- I'm in the building Halloween Jack lived on top of.
The Ziggy Stardust years were never my favorite.
ACTUALLY his persona on Diamond Dogs was slightly different from Ziggy Stardust. Halloween Jack has an eyepatch and a silk scarf and does NOT have a facial lightning bolt.
If anyone needs an earworm, I've been thinking Heroes off and on for a couple of weeks now. It's pretty good for that.
OK, let's listen again: https://www.youtube.com/watch?v=bsYp9q3QNaQ
114: I tried the package and got the warning also, and the warning appears in the documentation example without any explanation. So I think the software just always prints the warning, even for analyses with no weights. (The most common sense of "weight" is Moby's in 116).
After a logistic regression, the margins() function by default will estimate changes on the probability scale, not the odds ratio scale. The real-world impact is probably better estimated on the probability scale.
Probably probability.
But, Stata has given me that warning before. I don't remember what I was doing (and I know it wasn't margins).
OT: Is the occasional bout of crippling hip pain just something you need to plan on happening in your late 40s? I think I want a hip replacement, but don't think Med Express can do one before work tomorrow.
I'm afraid to take naproxen and drink at the same time and I can't untake the naproxen now that it didn't work.
There's going to be a special master to oversee the taint team. Best scandal ever.
Stormy Daniels and the taint team, all in the same courtroom. The mind boggles.
140: dang. I watched a lot of Night Court.
112. You should read "Scoop," by Evelyn Waugh.
I tried to read that and got too annoyed. I think I dropped it when he was at the outfitters. I've tried to read Waugh several times, but never got far.
Waugh was an evil son of a bitch who should have been strangled at birth. Even his friends hated him. Everybody can get so far and no further with his stuff. Some people baulk at Decline and Fall; some are OK with that and Vile Bodies but can't cope with Scoop and Black Mischief; Some fall at the hurdle of the war trilogy. Only fervent admirers of Boris Johnson can honestly say they enjoyed reading Bridehead Revisited. In the murky world of English conservative novelists, I'd give everything Waugh ever wrote for two more early Anthony Powells.
|| Double-ping for Londoners--I mentioned it back in 102 but it probably helps to do it at a time when you're all likely to be awake and procrastinating--to see if they still want to get together next week. |>
Maybe they just don't like you.
They're British, so they were trying to be tactful, but now you're forcing their hands.
Nope, I'm fleeing the country, Narnia-wards.
145: yes agreed, an unpleasant man and an author of unpleasant books...
They *really* don't like you.
re: 146
When are you around again? Sorry, my schedule is all mixed up, I've been away a lot for work. I might be around, though.
148: Absolutely. I intend to crush their social norms with wanton disregard like a bull in a china shop, because I'm American and goddammit that's what we do.
My books are perfectly pleasant, though.
152: This Friday (20th) until next Tuesday (24th), likely to be jetlagged on Friday but will probably be good for nighttime stuff.
145. How did you feel about "The Loved One"? TBH all I remember from "Scoop" is "Up to a point, Lord Copper"* and that it was set (mostly) in a faux-Ethiopia in the '30s.
* Which, for those who don't have the interest in actually reading the book, is how a certainly toady-ish character corrects his newspaper magnate boss when he says something utterly incorrect, such as "chris y is a big fan of Evelyn Waugh" or "elephants are native to Scotland." One of my friends has used it as a catch-phrase for decades, and the habit has affected (infected?) me as well.