(Thread IKs:
weg, Toxic Mental)
|
Man AI art loving sucks
|
#
?
Apr 28, 2024 18:50
|
|
|
|
| # ? May 29, 2024 14:35 |
|
|
Stink Billyums posted:he's still at it! History repeats itself. It's like the 60s. Communist infiltrators are agitating and causing the protests and now the police are getting called in to bust some heads. Let's hope it doesn't end up with the national guard showing up.
|
|
#
?
Apr 28, 2024 18:53
|
|
|
Grey Cat posted:Very big data set, mhmmm. About 1,000 is a common sample size for national-level polls. There may or may not be 99 other problems with this particular poll, but a small sample size isn't one of them.
|
|
#
?
Apr 28, 2024 18:55
|
|
|
Soros has been the Epstein of their Two Minute Hate for a long time now.
|
|
#
?
Apr 28, 2024 18:57
|
|
|
You mean Goldstein?
|
|
#
?
Apr 28, 2024 18:57
|
|
|
priznat posted:AI sucks
|
|
#
?
Apr 28, 2024 18:58
|
|
|
beer_war posted:About 1,000 is a common sample size for national-level polls. There may or may not be 99 other problems with this particular poll, but a small sample size isn't one of them. So, it's got 99 problems but it's reach ain't one?
|
|
#
?
Apr 28, 2024 18:58
|
|
|
Stink Billyums posted:update lol Two things I love about this: 1. She emphasizes that it was 20 years ago, despite making no suggestion that she wouldn't do the same thing tomorrow. 2. She points out that it wasn't literally illegal, as if the problem people have with her murdering her dog is that she might have been breaking the law. Just some amazing attempts at digging up.
|
|
#
?
Apr 28, 2024 19:00
|
|
|
beer_war posted:You mean Goldstein? Same diff, right?
|
|
#
?
Apr 28, 2024 19:00
|
|
|
"I'm the type of politician who's willing to learn from my mistakes. Also that wasn't a mistake, the dog was asking for it."
|
|
#
?
Apr 28, 2024 19:00
|
|
|
Stink Billyums posted:update lol can't even attempt to damage control her dog murder without plugging the obligatory book someone else wrote for her
|
|
#
?
Apr 28, 2024 19:07
|
|
|
BigglesSWE posted:https://twitter.com/atrupar/status/1784631051316232495?s=46&t=G1x8XWIwrNxUQoXItlkh2w On one hand RFKJR getting Ross Perot like numbers is baffling because Ross was way more reasonable than RFKJR, on the other hamd we've also had a couple more decades of paint chip eating, and microplastic huffing.
|
|
#
?
Apr 28, 2024 19:12
|
|
|
lol Stein pulling up the rear with a healthy 3%
|
|
#
?
Apr 28, 2024 19:16
|
|
|
Wingnut Ninja posted:Two things I love about this: The first and second sentence of her last paragraph are totally unrelated. Sure a dog that kills livestock can be put down. But she doesn't mention that. Only that Cricket "had shown aggressive behavior towards people" It's still illegal or at least was at the time.
|
|
#
?
Apr 28, 2024 19:16
|
|
|
priznat posted:Man AI art loving sucks I prompted AI to copy+paste an image of a poll doll that can do me 5-6 intimate polling results at a time and this is what I got.
|
|
#
?
Apr 28, 2024 19:19
|
|
|
RoyKeen posted:The first and second sentence of her last paragraph are totally unrelated. Sure a dog that kills livestock can be put down. But she doesn't mention that. Only that Cricket "had shown aggressive behavior towards people"
|
|
#
?
Apr 28, 2024 19:20
|
|
|
lol @ anyone who thinks RFK gets mod than like 3% nationally
|
|
#
?
Apr 28, 2024 19:22
|
|
|
Bottom Liner posted:The CNN poll was conducted by SSRS from April 18-23 among a random national sample of 1,212 adults drawn from a probability-based panel, including 967 registered voters. Surveys were either conducted online or by telephone with a live interviewer. Can someone unpack this for me? What is a probably-based panel? How do you have a 'national random sample' drawn from a 'probability-based panel'? Either it's a national poll, or its a poll of a sub set, not both. People keep talking like these samples are big enough, but like that only works if the sample is random. None of this seems random at all. Even if they have a random sample, and it's not drawn from a 'panel', they all seem to have some secret sauce corrective model to 'fix' the poll result, to correct for lack of young people and minorities, which they never release. I have had an issue with all these polls having Kennedy in double digits. How? Do even 10% of voters even know he is running?
|
|
#
?
Apr 28, 2024 19:22
|
|
|
WithoutTheFezOn posted:It’s also mentioned elsewhere that the dog killed someone's chickens. Fair enough. If she said that in that tweet it'd be easier to give her the benefit of the doubt. (a doubt?, I dunno)
|
|
#
?
Apr 28, 2024 19:23
|
|
|
Bone Crimes posted:Can someone unpack this for me? What is a probably-based panel? How do you have a 'national random sample' drawn from a 'probability-based panel'? Either it's a national poll, or its a poll of a sub set, not both. These pollsters design and publish polls that show all kinds of possible outcomes because the whole business model is manufacturing nail biting drama for mass media so CNN can say "BIDEN IS LOSING TO TRUMP BY 9 POINTS!!!"
|
|
#
?
Apr 28, 2024 19:25
|
|
|
The more insane you are, the more probable it is you will vote. That's the way to unpack it.
|
|
#
?
Apr 28, 2024 19:25
|
|
|
Oscar Wild posted:Kristi noem watches the first 10 minutes of John Wick and learns the wrong lesson Treating it the same way Drump treats "Bloodsport" and just fast forwards and rewinds to "the good parts"  Seth Pecksniff posted:Also it's Sunday go outside and play or something I can't, I conduct electricity too much and going outside is shaking hands with danger. 
|
|
#
?
Apr 28, 2024 19:26
|
|
|
BigglesSWE posted:https://twitter.com/atrupar/status/1784631051316232495?s=46&t=G1x8XWIwrNxUQoXItlkh2w the only way kennedy gets 16% in a poll is if they poll 6 people and he is one of them.
|
|
#
?
Apr 28, 2024 19:28
|
|
|
In response to the “any time, any place anywhere” debate demands, Biden should just keep choosing dates when Trump has to be in court. B: “Let’s have the debate next Thursday.” T: “Okay, yes….. um no I’ve just been informed that I have to be somewhere….”
|
|
#
?
Apr 28, 2024 19:30
|
|
|
Toxic Mental posted:lol @ anyone who thinks RFK gets mod than like 3% nationally Every nurse who lost their job will vote for him every nurse I know still has a job
|
|
#
?
Apr 28, 2024 19:30
|
|
|
redshirt posted:Soros has been the Epstein of their Two Minute Hate for a long time now. george has been retired for nearly a year now and isn't doing poo poo, i'm waiting for the right to cadaver synod him in perpetuity once he passes also w.r.t. vote shaming it's kinda funny how conservatives can do it all the time without complaints but it's always an issue on the left. gotta love that fall in love/fall in line dichotomy
|
|
#
?
Apr 28, 2024 19:33
|
|
|
BigglesSWE posted:https://twitter.com/atrupar/status/1784631051316232495?s=46&t=G1x8XWIwrNxUQoXItlkh2w So, basically, this poll had 25% of its sample size being people who are not even able to vote. Which actually seems pretty close to correct. https://www.statista.com/statistics/999919/share-people-registered-vote-age/ But if you're not even registered to vote, I fail to see why your opinion matters for the sake of these polls. This is probably why RFK and Trump have such high support in the poll. 25% of the poll is either "both sides suck," "I literally can't vote and have the political knowledge of an actual MAGA hat," or "I vaguely remember hearing the name Kennedy somewhere..." It's garbage.
|
|
#
?
Apr 28, 2024 19:34
|
|
|
Oscar Wild posted:On one hand RFKJR getting Ross Perot like numbers is baffling because Ross was way more reasonable than RFKJR, on the other hamd we've also had a couple more decades of paint chip eating, and microplastic huffing. The Boomers are full of lead. The Millenials are full of micro-plastics. By God, it's another glorious victory for Gen X. 
|
|
#
?
Apr 28, 2024 19:36
|
|
|
Deptfordx posted:The Boomers are full of lead. Gen X got both.
|
|
#
?
Apr 28, 2024 19:36
|
|
|
do all strong men pay aides to "Follow Him Around With a Wireless Printer to Print Out Good News From the Internet"? when i hear that poo poo i instantly think of characters like rambo, rocky and superman seems like something they would do
|
|
#
?
Apr 28, 2024 19:37
|
|
|
Mad Hamish posted:Ok well here's a Trump lol: that bit at the WH correspondent's dinner last night about Trump falling asleep in court and then farting himself awake. LIKE A DOG
|
|
#
?
Apr 28, 2024 19:38
|
|
|
blackmet posted:So, basically, this poll had 25% of its sample size being people who are not even able to vote. I appreciate the effort to actually look at the demos to infer some kind of structural problem with the poll, instead of just assuming it based on the top-lines. It's still HARKing (which is unavoidable here, not saying it's a fault of yours), but at least it's based on something that can be intelligently discussed.
|
|
#
?
Apr 28, 2024 19:40
|
|
|
blackmet posted:So, basically, this poll had 25% of its sample size being people who are not even able to vote. Where are you seeing this number?
|
|
#
?
Apr 28, 2024 19:40
|
|
|
DarkHorse posted:LIKE A DOG my dog has never farted so viciously as to wake itself up. it does wake ME up though...
|
|
#
?
Apr 28, 2024 19:41
|
|
|
blackmet posted:But if you're not even registered to vote, I fail to see why your opinion matters for the sake of these polls. Only registered voters were asked who they would vote for.
|
|
#
?
Apr 28, 2024 19:41
|
|
|
Micro-Lead
|
|
#
?
Apr 28, 2024 19:42
|
|
|
Stink Billyums posted:update lol It’s totally legal to kill dogs so it’s completely normal that I took our family pet outside and blew its head off because it mildly annoyed me.
|
|
#
?
Apr 28, 2024 19:44
|
|
|
Bone Crimes posted:Can someone unpack this for me? What is a probably-based panel? How do you have a 'national random sample' drawn from a 'probability-based panel'? Either it's a national poll, or its a poll of a sub set, not both. I’m guessing, without looking at what they actually say they do, that they have a pool of respondents who are nationally representative who are already signed up with the firm, and randomly sample from that. It’s a common way to reduce costs and avoid having to random digit dial and it’s various problems Weighting is correct and there’s nothing bad about it. Things are infinitely worse without weights they probably don’t know he’s running and are just filling in a box on the survey because they recognize the name Kennedy. That’s actually likely somewhat reflective of those peoples voting behavior if the election happened right now and they voted vOv
|
|
#
?
Apr 28, 2024 19:44
|
|
|
BigglesSWE posted:https://twitter.com/atrupar/status/1784631051316232495?s=46&t=G1x8XWIwrNxUQoXItlkh2w
|
|
#
?
Apr 28, 2024 19:47
|
|
|
|
| # ? May 29, 2024 14:35 |
|
|
Bone Crimes posted:Can someone unpack this for me? What is a probably-based panel? How do you have a 'national random sample' drawn from a 'probability-based panel'? Either it's a national poll, or its a poll of a sub set, not both. You're in the ballpark but statistics is just brain melting like that sometimes. So the first step is they build a panel, which is theoretically demographically identical or very close to that of what they are trying to model. This is really hard to do though for reasons you probably expect, it can be hard to find members of smaller demographics that are willing to participate in the first place. Depending on how they recruited the panel it could be a very significant way of throwing everything off down the line with compounding errors, which, given this has Jill Stein at 3% nationally is my initial hunch. In the context of political polling I'm guessing they struggle particularly with the following, let alone trying to find people who tick multiple for cross tabs and such:
From there you can sample randomly and it actually takes way fewer samples than you'd think for the sample mean to start to converge to the population mean (in this case the panel). Somehow laypeople have gotten it into their heads that anything less than like.... tens of thousands is a small sample size when really you start to get pretty close to the population mean with a hundred samples or less per the central limit theorem I alluded too previously. Larger sample sizes basically give you diminishing returns on how close your sample mean gets to the population mean. There are a couple implicit assumptions statistically that go along with that. Namely that your sampling is normally distributed, which I'm not going to check but nature loves a normal distribution so it's incredibly likely and given that its an industry standard technique I'm guessing it is. The other big one, and is another potential pitfall here, is that samples should be independent from each other. This actually may not necessarily be the case with voter opinions compared to a discrete, binary event like watching a TV show or not. Once you do all this, you have some weighting factor to scale your panel sample up to the population level. Basically this is just a big multiplying factor per person. Statistically everything they've done in that blurb is standard practice, it's what Neilsen has done for years to measure TV viewership. I actually worked on a project to emulate them a couple years ago. To give you a real world example of how easily this can go wrong, our initial panel beta testers were almost all people who worked for my previous large international manufacturing conglomerate that is mostly known in the west for electronics. We were massively over represented for middle aged, upper class white people because those are who works in ad tech. We had I want to say like.... ten users total in the initial rounds that were both younger and lower income. But we still have to scale that back up to the population, which means that one individual gets a much larger relative scaling factor. This lead to a couple cases where we would "find" things like every single zoomer watched Random Bullshit Show, but really all that happened was two of them watched it and our initial data was heavily skewed. I hope I've explained this well I really didn't want to effortpost about statistics and polling for a project that I've mostly tried to delete from my memory.
|
|
#
?
Apr 28, 2024 19:50
|
|
