|
DACK FAYDEN posted:edit: I clicked through looking for the data and not only could not find it but: Credibility intervals are a sort of Bayesian analog to frequentist confidence intervals (but with respect to the posterior distribution, not the sampling distribution). But I agree that their description is useless because it doesn't quantify how likely it is to be within 1%.
|
# ¿ May 1, 2018 21:56 |
|
|
# ¿ May 8, 2024 10:31 |
|
Jay Rust posted:With Canada: The people of Quebec have a culture and history which has evolved in a fundamentally different way from their French origin. In fact, what is surprising here is that the chart says "French Canadian" at all, given that a large proportion of people in Quebec do not consider themselves French OR Canadian; they are Quebecois. They are decidedly not "just Canadians who speak French". They've been trying to leave Canada for literally decades because of this. If you actually take a look at the results of Statcan's 2006 Census, you will find that "French Canadian" actually does not appear at all in the results, and that it was substituted instead of "Quebecois" in your chart. It's actually kind of offensive, to be honest.
|
# ¿ Jan 21, 2020 23:55 |
|
Jay Rust posted:Would a native French speaker from, say, New Brunswick call themselves or their heritage “Québécois”? The Cheshire Cat posted:No, from NB they would probably identify as Acadian. There was indeed a significant proportion of self-reported Acadians in that Census. There are no self-reported "French-Canadians". I'm very sorry that I forgot our Francophone minority friends in the other provinces. That's my bad.
|
# ¿ Jan 22, 2020 01:49 |
|
My favorite part is the outright lie in the description: nearly 3.3 million initial claims were not literally "filed" last week, that is a seasonally adjusted estimate. There were 2.9 million claims actually filed (almost 15% less), but they seem to be using a proportional seasonality adjustment, as if the seasonal effect applied uniformly to the entirety of the claims and not just to the typical 200k that would have lost their jobs that week regardless of the current crisis.
|
# ¿ Mar 27, 2020 12:54 |
|
Karia posted:Then there's the poly fit, which serves no value at all. Who cares what the R^2 is? You're not constructing a model of the data for predictive purposes, the fit doesn't matter. R^2 is of absolutely no use whatsoever, particularly for predictive purposes.
|
# ¿ Apr 17, 2020 21:37 |
|
Stoatbringer posted:"You! Unpaid lackey #17. You have an important mission. This is from THE VERY TOP!" Kevin Hassett is not an unpaid lackey, he is a professional clown whose entire career is abusing the data to make the numbers go up or down as needed.
|
# ¿ May 5, 2020 13:30 |
|
Uh there's nothing wrong with a line chart plotting something that's not "cumulative"? The issue is that they're using the phrase "V-shaped recovery" when this is actually closer to an "L-shaped recovery", but is shown in a way that it has a "V" shape. It's similar to the sometimes inconsistent usage of the phrase "flattening the curve".
|
# ¿ Jun 6, 2020 00:09 |
|
vyelkin posted:https://twitter.com/hereforthefin/status/1268928198508609537 I agree that the situation is very bad but that truncated y-axis on the "more realistic" chart is upsetting
|
# ¿ Jun 6, 2020 02:00 |
|
Platystemon posted:Do you also get upset by truncated axes on charts of temperature? Not sure what you're getting at here (I of course do get upset at all sorts of charts) but in your example not truncating the axis hides small variations which are of interest, whereas here it does not (this is an enormous drop and it would be visible even if you extended the y-axis to zero and showed the full history???). Truncating the axis here makes it seem like a "trick" to make the drop look larger, making it easy for idiots to attack you and call you a liar. It does not make the chart clearer whatsoever.
|
# ¿ Jun 6, 2020 02:27 |
|
"Life expectancy" is defined at birth and is not relevant for someone who is already 80. The expected remaining lifetime conditional on the current age is more relevant. There's a difference because death rates vary wildly over your lifetime (e.g. you are roughly 15x more likely to die before your first birthday than between the first and second). Based on some life tables I found in a quick search: in the US, at birth, a man can expect to live 76 years. However, if he reaches 80, he can expect to live to 88. So you're not "on borrowed time" and "should already be dead anyway" when you become 77, that's just a lie.
|
# ¿ Jul 8, 2021 00:01 |
|
Voyager I posted:Wouldn't 6% inflation functionally wipe out that GDP growth anyways? I'm assuming they don't factor that in to these calculations. Real GDP (i.e. adjusted for inflation) did indeed grow roughly 5.7% in 2021 as compared to 2020. GDP is almost always quoted in "real" (deflated) terms. Nominal GDP (i.e. not adjusted for inflation) grew 10%. It looks like a lot of "growth" (whatever that's supposed to mean to normal people), but current real GDP is still below its pre-COVID path, so it's not so much "booming" as "still catching up". You can see the data most easily from FRED, here. Look at the general trend leading up to 2020 versus where we're at now. ShimaTetsuo has a new favorite as of 03:41 on Jan 28, 2022 |
# ¿ Jan 28, 2022 03:39 |
|
Ajon posted:
This chart has been recreated from a 1994 article that committed the exact same dual y axis sin (I mean, it would have to, how else would you come up with the "crossover" term except by plotting it bad?). Can you imagine if "the terrible chart" was like the standard in your field, and decades later you constantly have to swallow your pride and reprint a version of it even though you know that it's wrong and bad but some moron decided it looked cool and you can't get anything published if you don't include it.
|
# ¿ Aug 4, 2022 12:57 |
|
lol at poo poo software that doesn't properly differentiate between zero and missing data double lol at the little bump between 2018 and 2019 caused by spline/smoothed line curve nonsense edit: that website posted:The current population of India is -1,617,838,620,736 as of Wednesday, September 28, 2022, based on Worldometer elaboration of the latest United Nations data. ShimaTetsuo has a new favorite as of 22:06 on Sep 28, 2022 |
# ¿ Sep 28, 2022 22:03 |
|
mobby_6kl posted:Wait what's going on here? Did the value overflow or something? 32 bits should be enough for like 2 billion I think. no clue but from looking at the projections for 2-3 other countries, looks like: - linear extrapolation based on the last two values (note: UN population data is for mid-year so the 2020 value is ~2.25 years ago) - that extrapolation is totally hosed by the bad zero in 2020 - some fuckery with the 000's and decimal separator and it should really be -1,617,838,620.736 not -1,617,838,620,736 (maybe???)
|
# ¿ Sep 28, 2022 23:35 |
|
Analytic Engine posted:gently caress that poo poo, I blame the impatient Product/Project Manager/Owner that's especially funny to me because UN WPP data is available through 2021 right now. so someone was in a great big hurry to poo poo out this website with zero Indian population but hasn't updated it for 2 whole years after that
|
# ¿ Sep 28, 2022 23:44 |
|
also "back WITH A VENGENCE" is a real weird thing to say for a projection that is really the same as in 2019 like it's the headline that goes with the "pent up demand" fairy tale of "well, people didn't go on vacation for 3 years so they're gonna go loving NUTS this year", but their projections are just kinda mid and they forgot to update the title
|
# ¿ Nov 16, 2022 03:01 |
|
it's a dumb chart but...(all-items) inflation *is* indeed down in the US. it's been down for about 6 months. that just means that prices are going up at a slower pace (not that prices are going down, or getting to anywhere near what they were before). my favorite part is that "supercore" is a subset of core, not a superset (like, it's "even more core"). economists are the dumbest people on the planet.
|
# ¿ Jan 22, 2023 16:16 |
|
Agnostalgia posted:A real competitor knows you always have to give 230%. you're looking at that wrong. The probabilities displayed are (G_i is the result of game i): P[winning championship | G_1, ..., G_10] And then: P[winning championship | G_1, ..., G_10, G_11=win] P[winning championship | G_1, ..., G_10, G_11=draw] P[winning championship | G_1, ..., G_10, G_11=loss] i.e. they're conditioned on different events, you can't just add them. (not a chess nerd so details are probably off but that's basically what it's saying)
|
# ¿ Apr 25, 2023 14:49 |
|
the constant talk of recession is not a forecast. it is a threat. it's corporations being their usual subtle selves: - threatening employees that they need to stop asking for raises so they can afford food, or else they'll get fired (if I raise wages then when was the point of raising prices???) - threatening the fed to stop raising interest rates (it's not fair to charge this much, you have to stop or i'll have no choice but to destroy the economy!!) also non farm payrolls get revised fairly dramatically every month: BLS posted:The change in total nonfarm payroll employment for February was revised down by so like 2 months ago, the story was that expectations were like +225k and actual was +326k, a "huge miss". but now the revised actual figure is just +248k, which is actually very close to those expectations. but we don't talk about 2 months ago anymore, we only talk about how this month the expectations were too low again! the financial press is a scam.
|
# ¿ May 5, 2023 22:40 |
|
Phanatic posted:How does that happen? Two points, which is what they have there, uniquely defines a line. How do they wind up with something so wrong? Equivalently to what DontMockMySmock pointed out, they are using three points. The third point is (0,0). However, it's weighted differently from the other two; in some sense it has "infinitely more" weight, such that any line that doesn't pass through it is considered worse than any that do. That's how you wind up with that.
|
# ¿ May 12, 2023 22:26 |
|
credburn posted:
That's an "event" (or a set, depending on the formalism). You can't add those together, so you'd never write "+ + S" or whatever, and it's pretty clear. But what the environmental story-telling is hinting at here is that it used to say: P(S|P) = P(S)P(P|S)/(P(S)P(P|S)+P(H)P(P|H)) and that was just too much, so they replaced P with +.
|
# ¿ Dec 4, 2023 03:40 |
|
Are any stores open on december 27th-30th? Is it missing because it's implied that it's normal hours? Or is it missing because it's implied that they're closed? It's probably the former, but how confident are you that Market Street specifically is actually open those days and not closed for most of the holiday season? It would also help to not present something that isn't a calendar (like with days in columns, weeks in rows) in a form where it looks just like a calendar at first glance. Not including the date on most of those days is also going to lead to confusion ("eve" vs "day"), as does having gaps between the days. Personally what bothers me the most is when you take up a lot of space to provide very little information like you were trying to fill up the space, but I'm still left with questions at the end of it.
|
# ¿ Dec 23, 2023 16:43 |
|
redleader posted:german looks like the output of one of those algorithms that can randomly distribute points evenly across a plane https://en.wikipedia.org/wiki/Low-discrepancy_sequence#Graphical_examples
|
# ¿ Jan 29, 2024 23:39 |
|
|
# ¿ May 8, 2024 10:31 |
|
Captain Hygiene posted:Elon thread found some quality graphs and charts content im the period as a thousands separator for numbers below 1 billion and then randomly switching to a comma above 1 billion
|
# ¿ Apr 25, 2024 20:27 |