Now this is an interesting argument, with graphs to back it up: the Gaza numbers aren't real.
Even if you take the numbers at face value, they put paid to the idea that this represents a 'genocide' by the Israelis; 30,000 is 0.2% of the Palestinian population, after four months of fairly intense urban warfare. If they really wanted to wipe out the 14MM people, they'd need to be working a lot harder at it than this. 99.8% of them are still alive, even if we accept the Gaza Health Ministry's numbers.
But we shouldn't, as the article lays out. The mathematical anomalies are such that the numbers look invented, not natural.
15 comments:
"Invented" numbers?
Like vote tallies in urban precincts? Like temperature readings in Siberia, Antarctica, and Greenland? Like employment / "jobs created" numbers from the US Dept of Labor? Like poll numbers from Gallup? Like estimates of incidence of chromosomal abnormalities resulting in natural phenotypical "trans-gender" conditions? Like analysis of the crime rates and tax contributions of pedestrians illegally entering the US across the southern border? Like predictions of the benefits of a taxpayer supported professional sports arena; or a convention center? Like forecasts of "excess death" from a contagious disease, or a rushed vaccine? Like standardized test scores in public schools? Like sales rankings for newly-released "Best Selling" books in the New York Times? For census data on Polar Bear populations on sea ice? Like heart attack rates for consumers of salt, or low dose aspirin, or red wine, or dark chocolate, or refined sugar, or second-hand smoke ...
In my youth I enacted for myself a rule of thumb that considered a newsworthy headline claim to be untrue if the authorities were unwilling to back the claim with a numerical context. In my now advanced years; I see a lot more numbers behind claims. But as far as I can tell, the percentage of "newsworthy headline claims" that, in a years' time or so, turn out to be untrue, is unchanged since I was about six.
It may be about to get worse, thanks to AI.
https://www.yahoo.com/news/proliferating-news-sites-spew-ai-012349209.html
But yes, if you look at almost any officially-backed narrative you'll find it to be at least partially false. The dietary items you close with are the ones that are most astonishing, because people had thousands of years of experience to go on when they started to be fed those stories. Belief in progress and science ended up being the enemy of health, as people proved to be willing to believe news stories and educational materials about 'the Food Pyramid' and suchlike. "It's healthier to eat this factory-baked white bread because it was cooked in an industrial oven that can get far hotter than your oven at home!"
Ethnic Cleasing
Greg
J Melcher - trying to kitchen sink that destroys your credibility. You can't just say "I think those are invented" because you don't like them. Some of them are fine.
This is what happens when anyone, not just you, tries to make a killer list to pound others to the center of the earth. Other stuff creeps in.
There is absolutely no data that is not skewed by desired goals. None. From an individual to the largest organizational body.
The only question is the degree of that bias.
The more to gain or lose, the bigger the bias, the more orchestrated the lies.
The biggest? Climate Change. Now there is an excuse to meddle in every facet of human behavior-nothing is exempt. It is the perfect tool for tyrants.
J. Melcher is far closer to the truth than comfort would allow.
I saw that the ACLU has turned on the National Labor Relations Board over the issue of arbitration, which it has normally opposed in favor of access to the courts, because it is itself involved in labor disputes it wants arbitrated.
https://www.nlrbedge.com/p/the-aclu-is-trying-to-destroy-the
Interest can definitely distort, and indeed is very likely to do so -- especially if there's a truth you want to believe, or a power you'd like to have that you would otherwise be denied. If it costs you money for anything else to be true, or it makes you money that this thing is true, it's hard not to have your vision distorted. It's very difficult for principles to triumph over such things.
It doesn't matter that the numbers aren't real.
The first article I saw questioning the Gaza Health Ministry's numbers was back in November, I think. The numbers have been bad the whole time. But people keep using the bad numbers because:
a) it fits the Big-Bad-Israel-Picking-On-the-Poor-Little-Helpless-Palestinians-Again narrative, and/or
b) there aren't any other numbers to use. No one else can offer a credible alternative and say "this is the real casualty count". In the absence of good data, we tend to default to next-best available data, even if we know it's flawed, because (in most cases) poor data is still better than no data at all.
Personally, I have dismissed casualty reports as nothing but propaganda. Obviously, people are getting killed and injured, but Hamas is a terrorist organization and I refuse to extend them any credibility - so when their puppet, the Gaza Health Ministry, says Israel has killed this many civilians, I dismiss it as a self-serving lie. (Either that or wishful thinking, considering their predilection for using civilians as human shields.)
There is absolutely no data that is not skewed by desired goals. None. From an individual to the largest organizational body.
Really? I don't think I agree.
To give an example, back when I was teaching full-time, I regularly tabulated data on test questions after giving midterms and finals to find most-missed and least-missed questions, etc. My goals were to use the results to guide test reviews and further instruction as well as point out possible flaws in my questions.
None of this was required by my employer, whom I don't think even knew I did this, and no one else ever saw the data or my analysis of it.
In what way do you think my data were biased by my desired goals?
In what way do you think my data were biased by my desired goals?
Not to answer for him, but probably by availability bias at least. Since you were thinking about the set of recently-asked questions, you weren't thinking about other things -- perhaps, for example, the way you used to ask the question a few years earlier; or the way your fellow professors would have framed the question, which was outside your data set because you framed the search as a private one.
I think the general point Raven is making that data is invariably skewed is probably right. It's something like Gödel's incompleteness theorems, or the recognition that you can't know both the position and velocity of a quantum particle. It's a good thing to recognize the limits of our models, even when we do the best we can with them. That doesn't mean we shouldn't use them, but it does mean we should humbly recognize our limits as human beings.
Hamas is a terrorist organization and I refuse to extend them any credibility....
I take a different view. Stipulate the numbers are accurate. That just emphasizes the evil that Hamas is perpetrating, since they're the ones responsible for the civilian casualties. Hamas is the ones using the Gazan civilians as shields, their residences as weapons caches and launching sites, their schools and hospitals as additional weapons caches and launching sites.
Israel, on the other hand, empirically demonstrable, publicly warns the Gazan civilians of impending strikes, thereby enabling them to leave before the strikes go in--and also allowing the targeted terrorists to escape before the strikes go in, except for those "martyrs" who stay behind to prevent as many civilians as they can from leaving, explicitly to run up the body count for the sake of terrorist supporter propaganda.
Eric Hines
I'm sure the gentleman is qualified and his analysis is accurate, and I'm in heated agreement with it (TM Steve Hayward). I don't think the article is particularly persuasive given the generally subjective nature of his comparisons. I wish examples of real casualty figures had been included rather than simple assertions of relationships even if the assertions seem grounded in reality.
It is probably better to simply note that Hamas has no interest in providing a factual accounting because it is in the business of justifying a brutal and largely unprovoked attack.
Grim, my knowledge of data work is limited, so maybe 'data' or 'bias' is being used here in a sense I'm not familiar with. Is it possible that by 'data' the interpretation of data is being included? I would agree that interpretation is fraught with all kinds of risks.
Yes, I completely agree that we have to do this with a clear understanding that there are limits, usually severe limits, to any data set.
However, I think I'm going to disagree with you on availability bias.* Let's take my example again. Since I'm not asking questions about past classes or about the classes of other instructors who teach the same class, data on them doesn't seem relevant. I am only asking about my current students' performance on a particular test, so the relevant data is there, and irrelevant data is not. (Nor do I draw conclusions about past classes or other instructors' classes from this data.)
I'm also only asking one real question (although I put it in different terms above): How many students missed each question?
That would seem to create an unbiased data set, although the interpretation of that data could well be biased or just wrong.
Or, let's take a marksmanship test for a rifle company. You have individual scores based on performance. That would also seem like an unbiased data set. The interpretation could be biased in various ways, but the data itself wouldn't seem to be.
All that said, yes, I do think all data sets have significant limits and interpretation can be quite difficult to do well, even in simple situations like my test results data. I could probably write a paper on all the things that could go wrong with interpretation of my test data sets, but I won't.
And, again, I am not a professional data wrangler, so maybe I'm just misunderstanding the claim.
###
*Okay, yes, if people want to claim that all assessments are inherently biased, which some scholars do claim, then that's their business. I don't really have much to say to them.
I think, Tom, that you may be interested in a different question than I am; this can lead to what is called 'talking past each other,' which isn't helpful.
So in the example of weapons shooting, there's a reason why we have long taught marksmanship to soldiers and Marines, and issued 'expert' badges or 'sharpshooter' badges according to the results. It's pretty objective: you shoot at targets at various ranges out to 300 meters, and we count the hits. 40/40 is a perfect score.
The USAF developed a slightly different scale that is meant to be equivalent. They shoot at a target at a single fixed range, which has multiple silhouettes that are bigger or smaller to adjust for the fact that they aren't actually 300m away. This is sometimes used outside the USAF for indoor ranges because it allows you to use such a range for a 'standardized' test.
The last time I qualified we used the USAF standard because we were shooting at an indoor range facility, which usually won't have 300m target capacity. I shot 39/40, which is well in the expert range.
Is that the same as shooting 39/40 on an outdoor range? No, of course not; there's no wind in an indoor range, just to name one difference (of which there are several others). It's much harder to complete the outdoor course than the indoor one.
On the other hand, one can argue, the indoor course may be a better test because it eliminates many factors outside the shooter's control (including the aforementioned wind). If you want to test for the shooter's actual ability to execute the technical task reliably, getting down to just the things in their control might be a better measure.
Back on the first hand, however, in a combat situation there probably will be many things outside of the shooter's control, and what you're really trying to teach is how to hit targets in that kind of environment.
So is the data useful? Sure. Is it one set of standardized data, or two sets? Does it matter? (Yes, for various reasons.)
Ultimately, though, this doesn't even get at the real issue. (cont.)
The real issue is that the test is always a substitute for something else. What you really want is effective soldiers and Marines, for whom the shooting is not a severable technical task nor one that can be replicated on an outdoor range either. We have chosen to test for this in the hope that it will improve something else, which is combat ability. There are some reasons to think that it does (for example, the relative effectiveness of early adopters like the US 3rd Infantry Division in WWI, when fighting against forces that did not receive similar training from Germany).
Whenever you're designing a test, then, you introduce bias in terms of what you choose to study. You have inevitably (which is Raven's point) introduced bias by selecting what you decided to test for and excluding what you didn't. This choice of what to measure is at least as significant as the choice of how to measure that thing (e.g., the USAF vs outdoor test).
I don't mean to suggest that it's not worth doing, just that there are unavoidable limits on our ability to use these tests and models.
You chose to test 'how many students got each question wrong.' There are other questions you could have asked; the choice to focus on that one was yours. The effect of the choice was to illuminate that part of the problem of student performance, making it something that you can study and track, at the expense of other things you could have done instead.
In larger-term projects, that kind of decision tends to have major effects downstream. We end up with teachers 'teaching to the test' because they're tracking that question -- which questions do students get wrong, because we want higher scores -- and not something else, which might be more important. Might be, but we will never know, because we can't study the issue -- there's no data, because we chose to seek this particular data instead.
On talking past each other, yes, that's why I kept wondering if I understood the terms.
I appreciate your explanations. I think that's good enough and I'll leave it at that.
Post a Comment