i don't feel like most people get "polls" and "studies"

[cursed knowledge]

like, the internet's full of this: "study founds that if you put a stick up your ass you get a 12% increased chance of dying at your 60s" and then people are like "OH MY GOD;;; ~~~SCIENCE~~~ JUST CONFIRMED THAT WE SHOULD PUT STICKS IN OUR ASSES"

and then there's this: "this online poll by shadysite.com said that 95% of the people actually love That Thing You Love Too™" and then you see all these old people saying "wow i knew i was right all along"

like, this isn't even exclusive to dumbass old gullible people, or stupid right-wingers

just a rant btw, people been stupid since forever, but i guess it's frustrating that almost everyone has a stupid view about "the science" and "the polls"

 

[Shelbyville Milhouse]

I ran a poll and 96% of people said you are wrong and I am right tho

 

[super-famicom]

Most people never studied statistics and it shows

 

[Stalker]

Studies show polling and studies are inconsistent and only useful for a general idea and should never be taken as fact.

 

[PlatypusDude]

Yeah, in addition to all the other stuff, the last four years have taught me that almost nobody knows how polls work. Like Aggregated National polling was closer to the actual margin in 2016 compared to 2012. Obviously some misses in how media portrayed them and some larger misses at state level, but folks on both sides of the political spectrum, flr differing reasons of course, take that as polls are flawed when evidence suggests otherwise.

So I guess part of the issues is most journalists are just as terrible as individuals in reading the polls. Gotta look for those who actually know what they are talking about.

 

[Hollywood Duo]

Most people can't do basic math let alone statistics and probability

 

[PeskyToaster]

I would say most people don't get much of anything.

 

[GMM]

Studies show that 69% of studies about how studies are misinterpreted

 

[Forerunner]

 

[Shelbyville Milhouse]

Also relevant

 

[Aaronrules380]

I definitely think it's a problem that we don't require statistics lessons in normal high school science curriculums given that literally any higher understand of science and scientific literacy depends so heavily on understanding statistics at at least a basic level. Both in terms of understanding what findings mean and if those findings are even useful or reliable in the first place

 

[ParanormalFupa]

People who criticize statistical analysis because it can't literally tell fortunes.....

 

[Studge]

The human brain did not evolve to easily understand things like statistics and large numbers so yeah, everyone is bad at it by default.

 

[mbpm]

XCOM has proven statistics is rigged against me

/s

 

[Trigonometrize]

*poll released that doesn't confirm your bias*

"Well they didn't poll me"

 

[Squarehard]

There's no poll or study to support this op, so I have no clue what you're talking about.

 

[ArkhamFantasy]

"Study shows oatmeal raises cholesterol just as much as hamburgers"

****Oatmeal was stuffed with lard, study funded by hamburger plant company***

 

[Rendering...]

No kidding, public education is shit.

 

[Karateka]

Always look for the p-value

 

[Manu]

Doesn't stop people from spoiling shit from you and then telling you without fail how this study says spoilers make you enjoy stuff better, actually.

 

[electroshockwave]

I've considered making a thread about some of this sort of thing here that bothers me. Often the results of a survey are posted and someone will disagree because they "only asked a few thousand people". But that number is a perfectly good sample size to apply the findings to the population, assuming there wasn't a selection bias.
Or the results of a study are posted about something like crime or discrimination etc and the first page is full of replies of "duh" or "they needed a study to find that out?". But even if the findings seem like something obvious, surely it's a good thing to have empirical evidence to point to? Especially if it's a social or political issue.

 

[MegaManTrigger]

Can confirm.

Had someone try to argue the following:

1) The second wave of Covid-19 was going to be weaker based on studies made on the H1N1 pandemic of 1918. In reality, they completely misinterpreted history and data regarding H1N1 actually being significantly more lethal in Wave 2. They also misinterpreted data in saying that mutations/Wave IIs are always weaker, when in reality, it's RNG as to whether the virus will weaken or strengthen in Wave II (The existence of D614G in the coronavirus RNA also lends credence to the fact that its virality is much greater now than before).

2) Had someone try to argue with me that masks weren't effective. His source? A study that said masks were singlehandedly one of the best ways to respond to a pandemic (his confusion in this case is that he misinterpreted the data on efficacy from low quality fabric masks to masks with filtration, and he thought that the low quality mask numbers meant that masks as a whole don't do shit).

 

[Weltall Zero]

I thought this would be about the infuriating "poll with a random sample of thousands is useless because actual population is in the millions" that Era members love to engage in.

Always look for the p-value

Click to shrink...

Indeed, but good luck getting forum members with a vested interest in disregarding the poll results to sit down and listen about what the null hypothesis or p-value are. Or that "correlation does not imply causality" is not a magic spell that excuses you from further questioning what the reason for the correlation is.

 

[Aaronrules380]

Always look for the p-value

Click to shrink...

To be fair, while p-values can matter, they can also be misleading themselves. Like a p value of 0.05 is usually the bar for what's considered signficant, but a significant difference in science doesn't mean a large one, just one we can be confident exists. A difference of 0.0000001 g on the amount of cocoa in a chocolate bar could technically be significant with enough tests, but it's still not actually important or relevant

 

[ArkhamFantasy]

The problem is that you have to look at each study or poll individually for what it is.

People aren't going to take the time to research who's funding this and look at the potential variables at play and any flaws in the nature of the study. That's why we get one study that says beef is good for you and another study that says beef will kill you.

 

[Chadtwo]

Wait so is your issue that you don't think people get polls or that they cite bad or fake polls?

 

[Maxim726x]

XCOM has proven statistics is rigged against me

/s

Click to shrink...

If anything, X-Com and other probability based games shows an important aspect of probability: 95% ain't 100%.

To be fair, while p-values can matter, they can also be misleading themselves. Like a p value of 0.05 is usually the bar for what's considered signficant, but a significant difference in science doesn't mean a large one, just one we can be confident exists. A difference of 0.0000001 g on the amount of cocoa in a chocolate bar could technically be significant with enough tests, but it's still not actually important or relevant

Click to shrink...

I think it's fair to be skeptical of medical research for this reason, and more importantly that the model used to plot the data can affect the P-value. Or, as I've seen personally, data can be 'massaged' to ensure that whatever variable is being studied arrives at a p value of <0.05.

 

[Forerunner]

XCOM has proven statistics is rigged against me

/s

Click to shrink...

MTG is rigged, all I draw are lands.

 

[Karateka]

Indeed, but good luck getting forum members with a vested interest in disregarding the poll results to sit down and listen about what the null hypothesis or p-value are. Or that "correlation does not imply causality" is not a magic spell that excuses you from further questioning what the reason for the correlation is.

Click to shrink...

Yeah I hate this tendency.
People need to realize that if something is unethical to experiment on correlation can be the best thing researchers have to go on... Luckily most are not going off with stuff like "correlation does not imply causation so smoking doesnt cause cancer"
Still most dont consider the amount of statistical analysis going on.

If anything, X-Com and other probability based games shows an important aspect of probability: 95% ain't 100%.







I think it's fair to be skeptical of medical research for this reason, and more importantly that the model used to plot the data can affect the P-value. Or, as I've seen personally, data can be 'massaged' to ensure that whatever variable is being studied arrives at a p value of <0.05.

Click to shrink...

Yeah but you can probably tell this if you read the study and the section on statistical analysis of the results.

If you just read the news reports they wont talk about the actual numbers anyways.

 

[Pau]

I definitely think it's a problem that we don't require statistics lessons in normal high school science curriculums given that literally any higher understand of science and scientific literacy depends so heavily on understanding statistics at at least a basic level. Both in terms of understanding what findings mean and if those findings are even useful or reliable in the first place

Click to shrink...

Agreed 100%.

Always look for the p-value

Click to shrink...

Eh, if anything introductory statistic courses focus on p-values too much (or rather, with the wrong emphasis) to the point that it results in researchers applying it blindly despite the data not meeting the necessary assumptions. Although this might differ across disciplines. It's definitely a problem in the social sciences...

 

[Elfotografoalocado]

I think everything relevant has been said at this point, people know nothing about causality, uncertainty, experimental designs, sampling, the infinite amount of things that can qualify a conclusion, statistics... Without intense training, people think in simple causal terms, in stereotypes, confirming their own biases... Hell, even scientists often do it.

 

[Bishop89]