>> found that 62% of respondents reported no or limited statistical knowledge
The other 38% didn't understand the question...(my extrapolation)
I did a couple semesters of statistics at uni. And I can confidently say that the number of people who can answer 3 simple questions on statistics (like say mean versus medians, confidence levels or margins of error) is, well, a rounding error from 0.
Indeed, statistically, no-one has a clue how statistics work.
I did however learn enough to know that statistics can tell you absolutely anything you want them to say. Assuming you don't just make them up, they're trivial to manipulate to generate the headline you want.
When used to evaluate risk, the comprehension goes down further (a fact willfully exploited by any decent marketing.)
One of my favourite jokes: 93 % of statistics are made up on the spot, and 61 % of people believe them.
Bonus points for changing the figures every time you retell the joke.
You _have_ to change the numbers each time, because you're constantly measuring and these things have basic variation. It would be unrealistic if the fake statistics were the same every time!
I’d recommend reading “How to make the world count” instead of a work of a morally bankrupt guy who used the very same tricks he criticized to discredit the cancer studies of tobacco usage.
> mean versus medians, confidence levels or margins of error
It's basic when you are attending an undergraduate course but most people can understand mean (as a dictionary might generically define it) and have a general feeling for margin of error (again, not in the mathematical way.)
Statistics has been the most difficult course in my CS course. For some reason when I start counting events to get a probability I find several perfectly plausible ways to count them, get five different probabilities and none of them is the correct answer.
And about being "trivial to manipulate [numbers] to generate the headline you want" a politician once told me that you can show the same numbers in any way you want, as in to demonstrate a thesis or its opposite.
I started thinking about this percentage on decision makers around the world and got a chuckle.
I've been trying to talk about the median vs. mean with a bunch of politicians on themes around the zillionaires, wealth or consumption and for most parts they're clueless. Or the ones with degrees still go with the normal (mean that skews the normal) as they're afraid.
I just happened to get a copy of naked statistics yesterday, as I feel the human traits of poor comprehension of probabilities can be enhanced to at least some extent. And my degree from the social side didn't include statistics.
If there are better entry-level books on the matter I'm happy to take some recommendations.
I wish people had an intuitive understanding of probabilities. It seems the average person can only think in terms of "basically never happens", "fifty fifty" and "sure thing".
It would be interesting to probe the political orientation of the people who understand and those who don’t and see if there is some correlation there.
My thesis is that no-one understands. So there's no political orientation.
Your suggestion suggests that a little bit of technical education would change political allegiance. Alas that is incorrect. Politics is about worldview (me and mine versus you and yours) and a world view cannot be changed by something as mundane as education.
Each person has both a self-centered side, and a community-centered side. For some it's mostly self centered. For others mostly community centered. Politics is about finding out which side the population has swung to.
Exactly. 38% is way too high. Are we sure about that? Statistics is not some required learning in school. Even those who learned (me for example) cannot confidently claim what it actually means. I would say most of the people don't have a clue what statistics mean
i mean, the headline statistic can be misleading, but you gotta dig in and wrestle with the details. just like anything, we cannot boil down complex things to single numbers and expect any sort of meaningful signal. we gotta roll up our sleeves, look at definitions, think about what our actual questions are, how we might answer those questions through measurements and observations, and what the confounders are. i think a common issue folks have with stats is that they expect a tidy answer, and it just doesn't do that: it's more of a way to prove the world...the results still need some interpretation.
The fact that 62% of Americans have little or no understanding of statistics may be related to those 40% of Americans that believe in Creationism, i.e. human (and fossils) were created by God a few thousands of years ago, and not of randomness and evolution over millions of years. I think no other industrial country has that disbelief in science.
https://news.gallup.com/poll/261680/americans-believe-creati...
This problem with science is apparent in another survey: in 2009, a Pew Center publication showed that 33% of scientists in the USA believed in God (and 18% in a transient power), which is much lower than the 80% belief of the general American population at the time. Of course, this is not a proof of causality in either direction, but scientific knowledge is seemingly inversely correlated to religiosity. And the USA are still more religious than any other industrial more-or-less-democratic country.
I'd say in reality it's way more than that. The first statistics course in uni was a very humbling experience. I realised that while I thought I understood a lot (and I was coming from a CS heavy background, olympiads and such) real statistics is way harder and a lot more counterintuitive than I thought. Granted, this talks about "basic" statistical understanding, but even that is way more complicated than most people assume.
yeah, i've studied a lot of math and a lot of cs, and stats is tough. part of the problem is the terminology, and just giving a ton of complex machinery without telling you what it's actually doing. i've never learned that way, and it is very easy to feel like you're doing some sort of dark magic.
also, probability theory vs statistics is an important distinction: prob theory is a nice clean mathematical subject, while statistics is almost the philosophy of applying probability theory to the world.
Statistical reasoning isn't really motivated when it's taught, at least in the US. My schooling (I did jump around a bunch) assumed the student to have picked it up through vague balls-and-bins style problems taught in various units in various grade levels.
By the time I took my statistics class in undergrad math, coming from a similar background to you, they just sort of assumed you had a head for combinatorics and used that to develop everything else. I was a really good student in undergrad and statistics was my hardest class, I spent like 2x time on that class than any other class.
In grad school I took a class on complex system failure analysis and was quite apprehensive. My hope was that I could team up with a classmate to help with the math while I could work on the systems levels analysis. Turns out that because I understood systems really well, system failure offered me the intuition I needed to really understand statistics. I aced the class, published a moderately popular paper in distributed systems using what I learned, then went and took our graduate level statistics class widely known to be very difficult and aced it.
I think tacking statistical thinking on as an afterthought in curriculum is a huge mistake in the school system, especially so in the age of machine learning. I think for the average student statistical thinking is even more important than a lot of trigonometry.
Even leaving aside the quantitative stuff (p-values, medians, whatever) and can't crunch the math, IMHO at least you should have seen how statistics can lead you to the exact opposite conclusion from reality, so that you at least know whether to think twice about a conclusion drawn from statistics thrown at you. Yet I was recently quite surprised to learn that even many folks in tech had never encountered Simpson's paradox before. All it takes to start explaining that is a scatterplot and a few lines, and yet it doesn't seem to be taught widely. It's rather terrifying, given that most people (myself often included) will be happy to believe "obvious" conclusions drawn from seeing one percentage greatly exceed another.
This was abundantly clear when people, even on HN, were upset about the bureau of labor statistics revising their numbers tendentially downwards, probably confusing the notion of statistical bias for that of political bias.
If the figures are biased, just estimate the bias and correct for that, what is the big deal they said, as if the bias variance tradeoff was not a thing.
There is the old joke that 87 percent of all statistics is made up on the spot. I’ve told it many times but a fair amount of people seemed to believe it hook, line, and sinker.
I would be more likely to believe the results if they tested these adults and not asked them.
There's a big ego hit in admitting you don't know something. And many people are brought up thinking that it's a shame not to know something and that someone is better for knowing something. Like, a better person, not just better in some field.
A coarse understanding like "the smaller p-value is the more likely a headline is true" is worse than no understanding at all. I bet most people who believed they understood what p-value is are like that though.
> The survey showed that 62% of U.S. adults self-report having little to no idea what statistics or statistical concepts like p-values are, but 90% of them would base decisions on reported statistics at least sometimes if they understood them better.
You're kidding, right? 38% self-report more than that? If their self-report were accurate it would imply an education system that has truly excelled.
> How much do you understand about statistics and p-values?
I suspect that it’s far far less than 38% of people who actually understand statistics to this level. I suspect if someone on HN went around and asked their co workers to explain what a P value is in 2 sentences, it would be less than 10% of a (presumably) highly educated workforce.
I suspect about 40% of adults are unable to tell the difference between mean/median/ mode, or could answer the Monty hall problem, or even “if I flip a coin 3 times are the chances I get heads 3 times in a row”
Most HNers don't even understand what percentages are and throw out dumb statements like "mega corp can treat 1% of users like crap, it's a small number!"
I mean, the Monty Hall problem is a literal gotcha that trips up literal professors, that's a horrendous example to use as the baseline for "basic understanding of statistics".
Other than that choice of example, I do agree in that I doubt anywhere near 40% of adults have basic statistical literacy. I've played in card game tournaments semi-professionally and just gambler's fallacy + results-oriented thinking alone make it so easy to take other people's money, and if you can't figure out such basic concepts as "getting tails once doesn't mean I'm due for a heads next flip" even when you're literally losing money, what are the chances of anyone else caring about understanding it when they're not even being given the hands-on reward-based reinforcement learning opportunity?
Yeah I maybe should have ignored the Monty hall problem - although I’d guess if you’ve studied enough stats to know what a P value is, and how to measure it, you’ve come up against the Monty hall problem!
What does it really mean to properly understand p-value? What I remember is that if the p-value is less than 0.05, the research result is considered statistically significant. That's about as far as my memory goes. I know that's actually a misunderstanding, but that's how most people understand it. I'm not sure how much I need to know to say I truly understand it.
It is mumbo jumbo (also people hearing “significant” treat this as “effect/change is large and important” which is in no relation to the actual amount of change).
Low p-value basically means how surprising your data would be if there were actually no effect (ie less than 5% of the time you’ll get this due to randomness if there is no change — which is rather impossible)
Sample size matters heavily. With more observations, estimates become more precise, so increasingly small differences can become statistically significant. With a large sample, you can therefore get a tiny, practically meaningless effect with a very small p-value.
Eg effect of $1 can be statistically significant (not random) which does not matter in practical terms if average is like $10000.
So the key point here is not only to look at the p-value but also at an actual change. If a drug gives you only 0.01% more hair, it doesn’t matter to you that it is guaranteed.
Agree. I think it's more important to understand statistical fallacies (selection bias, regression to the mean, survivorship bias, etc). Those are extremely common trip wires but to recognize them you don't need to memorize formal definitions.
Fair play to you for answering, but that’s not what a P value is! A p value is the chance of getting a result greater than the result you actually got, _assuming you are testing a hypothesis_.
I’m also pretty sure I would fall in the camp of saying “nope don’t understand P values” as I can’t remember anything else about them.
Same. I think it involves doing some other stuff very much properly... Like formulating whole thing... And even then you might luck out if you try enough of things...
So I admit I only know that P values are somewhat useful some of the time.
P-values are almost always taught poorly, but it's not actually that difficult of a concept. I took statistics in high school, again in undergrad, and it wasn't until the third time in grad school that it actually made intuitive sense (thank you Julia Yang!). When you're testing a hypothesis in statistics, it's easier to formulate a "null hypothesis" which is the opposite of what you're testing, and then try to disprove that null hypothesis.
A p-value is the probability, assuming the null hypothesis is true, of obtaining a result at least as extreme as the one actually observed.
Put differently: if the null hypothesis were true, then for p=0.05 you'd see <things at least as far from the test statistic as what you just observed> at most 5% of the time.
Put differently again: If the null hypothesis you are testing is true, then for p=0.05 random sampling would not return an observation as far from the test statistic as you just observed, 95% of the time.
Stats is one of those subjects which displays the dunning kruger effect really well.
Sure on the surface stuffile averages, medians, standard deviations etc are quite easy but the moment you start going deeper you realize how difficult and un-intuitive things start to get.
I'm still in the valley of despair, and frankly it's probably the best place to be in for the average person. Understand the basics and know just enough to realize how easy it is to mislead people with it.
I can beat a gorilla in a fist fight because it's gonna rip my body in half, but lose by disqualification due to rules violations. A posthumous win is still a win.
I think the average person is definitely intelligent. It's just that we are very prone to mixing intelligence with knowledge in addition to also forgetting how long it took us to learn things that are so automatic for us we see them as trivial.
Given that most measures of intelligence follow nearly normal distributions (as long as you stay away from the tails), I think it doesn’t matter that much.
Statistical literacy, or any literacy for that matter, requires hard work that is seen as a normal way of life. When machines do that hard work for us, it is no longer seen as a normal thing, but some unnecessary hard work (when we have cars, why walk 20 miles? And we have machines that do statistics).
Just like how those college kids saw my work as horrifically weird hard work, while I saw it as a normal thing.
My interpretation is how the risks (statistics) of those job functions are more known, people choose to avoid them. But the people that are-the-statistic don't necessarily agree with or understand those statistics.
For example, if OP is a coal miner that hasn't had health issues yet, they may choose to discount statistics that declare x% of coal miners have negative health outcomes.
Nope. It's a true story. And how it relates to thread is given below in another of my comments. "Things change with time" should have given you some relation. Statistics is seen by the current generation as hard work, rightly so, due to availability of easier ways of dealing with it.
Oh we used to dream of only walking 20 miles a day.
We used to get up at 3am half an hour before we went to bed, eat a lump of cold poison then walk FOURTY miles uphill to school then when we got home our Dad would slice us in two with a bread knife
The other 38% didn't understand the question...(my extrapolation)
I did a couple semesters of statistics at uni. And I can confidently say that the number of people who can answer 3 simple questions on statistics (like say mean versus medians, confidence levels or margins of error) is, well, a rounding error from 0.
Indeed, statistically, no-one has a clue how statistics work.
I did however learn enough to know that statistics can tell you absolutely anything you want them to say. Assuming you don't just make them up, they're trivial to manipulate to generate the headline you want.
When used to evaluate risk, the comprehension goes down further (a fact willfully exploited by any decent marketing.)
Statistically, most statistics are meaningless.
well encapsulated in the quote popularized by Mark Twain "Lies, damned lies, and statistics" [1]
[1] https://en.wikipedia.org/wiki/Lies,_damned_lies,_and_statist...
One of my favourite jokes: 93 % of statistics are made up on the spot, and 61 % of people believe them. Bonus points for changing the figures every time you retell the joke.
A statistician is a man who with his head in the freezer and his feet in the oven can say "On the whole I feel perfectly normal."
Recommended reading: 'How to lie with statistics' - Huff, 1954
:)
Our stat prof was a very special guy.
If even one person has 1 leg, then the average is strictly less than 2.
It's basic when you are attending an undergraduate course but most people can understand mean (as a dictionary might generically define it) and have a general feeling for margin of error (again, not in the mathematical way.)
Statistics has been the most difficult course in my CS course. For some reason when I start counting events to get a probability I find several perfectly plausible ways to count them, get five different probabilities and none of them is the correct answer.
And about being "trivial to manipulate [numbers] to generate the headline you want" a politician once told me that you can show the same numbers in any way you want, as in to demonstrate a thesis or its opposite.
Rhetoric teaches how to argue with words. Statistics is the same, but with numbers.
I've been trying to talk about the median vs. mean with a bunch of politicians on themes around the zillionaires, wealth or consumption and for most parts they're clueless. Or the ones with degrees still go with the normal (mean that skews the normal) as they're afraid.
I just happened to get a copy of naked statistics yesterday, as I feel the human traits of poor comprehension of probabilities can be enhanced to at least some extent. And my degree from the social side didn't include statistics.
If there are better entry-level books on the matter I'm happy to take some recommendations.
That's bonkers. Society is all about statistics, and the reverse. Without statistics any kind of social studies is only Just So stories.
Your suggestion suggests that a little bit of technical education would change political allegiance. Alas that is incorrect. Politics is about worldview (me and mine versus you and yours) and a world view cannot be changed by something as mundane as education.
Each person has both a self-centered side, and a community-centered side. For some it's mostly self centered. For others mostly community centered. Politics is about finding out which side the population has swung to.
When told to a statistically illiterate person who isn't aware of Simpson's paradox and so on?
Which is a rounding error from 100% according to the GP.
There's a reason actuaries get paid the big bucks.
Combined with Dunning-Kruger, this means that the real number of people completely clueless about statistics is closer to 38%.
[0] https://pmc.ncbi.nlm.nih.gov/articles/PMC8992690/
[1] https://www.sciencedirect.com/science/article/abs/pii/S01602...
This problem with science is apparent in another survey: in 2009, a Pew Center publication showed that 33% of scientists in the USA believed in God (and 18% in a transient power), which is much lower than the 80% belief of the general American population at the time. Of course, this is not a proof of causality in either direction, but scientific knowledge is seemingly inversely correlated to religiosity. And the USA are still more religious than any other industrial more-or-less-democratic country.
also, probability theory vs statistics is an important distinction: prob theory is a nice clean mathematical subject, while statistics is almost the philosophy of applying probability theory to the world.
By the time I took my statistics class in undergrad math, coming from a similar background to you, they just sort of assumed you had a head for combinatorics and used that to develop everything else. I was a really good student in undergrad and statistics was my hardest class, I spent like 2x time on that class than any other class.
In grad school I took a class on complex system failure analysis and was quite apprehensive. My hope was that I could team up with a classmate to help with the math while I could work on the systems levels analysis. Turns out that because I understood systems really well, system failure offered me the intuition I needed to really understand statistics. I aced the class, published a moderately popular paper in distributed systems using what I learned, then went and took our graduate level statistics class widely known to be very difficult and aced it.
I think tacking statistical thinking on as an afterthought in curriculum is a huge mistake in the school system, especially so in the age of machine learning. I think for the average student statistical thinking is even more important than a lot of trigonometry.
More than half would not be able to answer this, and not only in US.
100% of headlines of statistics-related articles must follow this rule.
Yeaaah… let’s talk about that.
Seems like a fair bit of stats were designed to intimidate —so as to get people to stop asking questions. Or at least that is the effect!
Stats designed for intuition are few. See Kill Math for how it might be done: https://worrydream.com/KillMath/
https://vimeo.com/906418692
https://worrydream.com/LadderOfAbstraction/
If the figures are biased, just estimate the bias and correct for that, what is the big deal they said, as if the bias variance tradeoff was not a thing.
I have seen too many parents (i am a teacher) lacking basic mathematical and geometrical understanding.
Let alone, statistics.
Schooling is broken, society is rotting, culture is dead.
There's a big ego hit in admitting you don't know something. And many people are brought up thinking that it's a shame not to know something and that someone is better for knowing something. Like, a better person, not just better in some field.
"You know how dumb an average American is. Well, mathematically speaking, half of them are even dumber than that."
No surprise here.
But if they release headline “62% of respondents” reported no or limited statistical knowledge while only 11% regularly use statistics in daily life…
…then they would loose more than half of readers who don’t know “per cent” or % symbol (?)
I thought that OP changed the original headline but no, psu.edu really published this :)
You're kidding, right? 38% self-report more than that? If their self-report were accurate it would imply an education system that has truly excelled.
I suspect that it’s far far less than 38% of people who actually understand statistics to this level. I suspect if someone on HN went around and asked their co workers to explain what a P value is in 2 sentences, it would be less than 10% of a (presumably) highly educated workforce.
I suspect about 40% of adults are unable to tell the difference between mean/median/ mode, or could answer the Monty hall problem, or even “if I flip a coin 3 times are the chances I get heads 3 times in a row”
Other than that choice of example, I do agree in that I doubt anywhere near 40% of adults have basic statistical literacy. I've played in card game tournaments semi-professionally and just gambler's fallacy + results-oriented thinking alone make it so easy to take other people's money, and if you can't figure out such basic concepts as "getting tails once doesn't mean I'm due for a heads next flip" even when you're literally losing money, what are the chances of anyone else caring about understanding it when they're not even being given the hands-on reward-based reinforcement learning opportunity?
Low p-value basically means how surprising your data would be if there were actually no effect (ie less than 5% of the time you’ll get this due to randomness if there is no change — which is rather impossible)
Sample size matters heavily. With more observations, estimates become more precise, so increasingly small differences can become statistically significant. With a large sample, you can therefore get a tiny, practically meaningless effect with a very small p-value.
Eg effect of $1 can be statistically significant (not random) which does not matter in practical terms if average is like $10000.
So the key point here is not only to look at the p-value but also at an actual change. If a drug gives you only 0.01% more hair, it doesn’t matter to you that it is guaranteed.
I’m also pretty sure I would fall in the camp of saying “nope don’t understand P values” as I can’t remember anything else about them.
So I admit I only know that P values are somewhat useful some of the time.
A p-value is the probability, assuming the null hypothesis is true, of obtaining a result at least as extreme as the one actually observed.
Put differently: if the null hypothesis were true, then for p=0.05 you'd see <things at least as far from the test statistic as what you just observed> at most 5% of the time.
Put differently again: If the null hypothesis you are testing is true, then for p=0.05 random sampling would not return an observation as far from the test statistic as you just observed, 95% of the time.
https://en.wikipedia.org/wiki/Misuse_of_p-values
Is that a well designed survey question?
I would expect the use of a specific jargon term in that question to affect the results in a significant way.
Sure on the surface stuffile averages, medians, standard deviations etc are quite easy but the moment you start going deeper you realize how difficult and un-intuitive things start to get.
I'm still in the valley of despair, and frankly it's probably the best place to be in for the average person. Understand the basics and know just enough to realize how easy it is to mislead people with it.
On a somewhat related note, 8% of Americans say they can beat a gorilla in a fist-fight.
Just like how those college kids saw my work as horrifically weird hard work, while I saw it as a normal thing.
For example, if OP is a coal miner that hasn't had health issues yet, they may choose to discount statistics that declare x% of coal miners have negative health outcomes.
We used to get up at 3am half an hour before we went to bed, eat a lump of cold poison then walk FOURTY miles uphill to school then when we got home our Dad would slice us in two with a bread knife