How Not to Fall for Bad Statistics - with Jennifer Rogers
ฝัง
- เผยแพร่เมื่อ 18 พ.ค. 2024
- Living is a risky business. If you believe the headlines, bacon is as deadly as smoking and fizzy drinks make children violent, but is that true?
Subscribe for regular science videos: bit.ly/RiSubscRibe
From causation and correlation, to relative and absolute risk, Jennifer Rogers explains how to figure out if the stats we are presented in newspapers are accurate.
Jennifer Rogers holds the position of Director of Statistical Consultancy Services at the University of Oxford having previously worked as a Post-Doctoral Research Fellow in the Department of Statistics funded by the National Institute of Health Research. She has a special interest in the development and application of novel statistical methodologies, particularly in medicine. Her main area of expertise is the analysis of recurrent events and her research has recently focused on developing and implementing appropriate methodology for the analysis of repeat hospitalisations in patients with heart failure but her research has many other applications in medicine such as epilepsy and cancer, but also in retail and engineering. She works alongside other statisticians, clinicians, computer scientists, industry experts and regulators.
This talk was filmed in the Ri on 11 March 2019.
---
A very special thank you to our Patreon supporters who help make these videos happen, especially:
Dave Ostler, David Lindo, David Schick, Erik Shepherd, Greg Nagel, Ivan Korolev, Joe Godenzi, Julia Stone, Lasse T. Stendan, Lester Su, Osian Gwyn Williams, Paul Brown, Radu Tizu, Rebecca Pan, Robert Hillier, Robert Reinecke and Roger Baker.
---
The Ri is on Patreon: / theroyalinstitution
and Twitter: / ri_science
and Facebook: / royalinstitution
and Tumblr: / ri-science
Our editorial policy: www.rigb.org/home/editorial-po...
Subscribe for the latest science videos: bit.ly/RiNewsletter
Product links on this page may be affiliate links which means it won't cost you any extra but we may earn a small commission if you decide to purchase through the link. - วิทยาศาสตร์และเทคโนโลยี
![คบซ้อน - วงL.กฮ. [ Official MV ]](http://i.ytimg.com/vi/5MtesqJ0I1c/mqdefault.jpg)
It is really nice to have finally arrived at point in British societal history, where an academic hasn't felt the need to abandon their regional accent. Another great RI lecture, made by another talented mind.
I actually watched this whole lecture primarily to listen to her accent, which I found quite charming. I will say that I didn't learn much, mostly because little of this was new to me.
Are you kidding me? Nobody is allowed to commentate on YT anymore without a British accent... I haven't heard anybody pronounce an "R" in AGES.
@@mikaelgaiason688 She doesn't just have a "British" accent, but a very distinct regional accent. Lancastrian, if I'm not mistaken.
I hear Coronation St.
7:05 "Statistical significance just tells us whether or not something definitely does or definitely doesn't cause cancer." I believe this should have been worded quite a bit differently...
Depending on the significance you either decide yes or no (accept or reject the null hypothesis). There is no partial yes or partial no.
James Lopez i agree - the word definitely cannot be used because there is always a risk that you are wrong - in fact the Type 1 error of the test is the risk. If you set the test at 5%, it means that there is a 5% chance that
If you reject the null hypothesis, you are wrong in doing so.
@@MN-sc9qs You do decide yes or no, but science is built on inductive reasoning, not deductive. Statistical significance means your results are very unlikely to have been due to chance, not "definitely not due to chance". Also, in the context of what was being referred to at the time (carcinogens) almost none of those studies are even designed to show causal relationships. The studies she was referring to show that it's very unlikely that X carcinogen is linked with cancer due to chance. Most of the time, "very unlikely" means 5% chance.
That sentence is what sparked a series of twitter discussions among stats. You definitely can’t use the word “definitely” and I’d even avoid “X causes Y”.
Ugh, this is one of the most important and needed lectures on this channel and it has less than 2k views. Almost everyone can learn from this. It's relevant to daily life. So many stats get thrown around these days by players just to push their political agenda and/or ideology, in order to give the impression that it's grounded in reality and supported by scientific fact. People fall for it time and again!
As a girl who took an extra curricular interest in statistical analysis since middle school, it's deeply frustrating to me.
Great lecture!
Let's see if it worked: Did you freak out because of Covid, did you insist everyone around you wear a mask?
Covid was one big hysteria freakshow. Hope you were able to keep your head on straight.
I had just finished eating cheese right before she started the "spurious statistics". I was brushing my teeth during that segment and now I'm going to bed. Wish me me luck!!!!!!!
10 months later and he still hasn't been heard from. 🙏
A favourite phrase of mine: lets play "abusing stats to prove a point".
David Young or lets torture the data so it tells the story we want to hear!
97% of stats will agree with me anyway, so why worry about the other 4%?
;)
@@bernardmcgarvey4169 "...so that IT tellS the..." - Oh dear! See my comment above.
If I'd seen this video when I was a kid, there's an 87% chance I would've majored in statistics in college.
www.xkcd.com/552/
And if you'd not seen it as a kid, there's a 13% chance you wouldn't have majored in it!
And a 100% chance you wouldn't have done either if you'd never been a kid!?!??
One of my favorite recent abuses of stats is "in the USA, the average Attorney makes the same money as the average Doctor" This conveys nothing about the relative incomes of the two groups. Most Doctors make within a narrow range of each other, near the average Doctors salary. Attorney's salaries widely vary from almost penniless to multimillions per year. More useful perhaps would be the median income and the standard deviation.
That's why we have "medians", standard deviations and interquartile ranges....
Also got to say this is to date my favorite Royal Institute video she did an awesome job and communicated something very important and relevant
Excellent stuff. Always be sceptical of news stories that include stats. Be doubly sceptical of news headlines that are questions rather than statements. Be triply sceptical of "9 out of 10 cats"!
Betteridge's Law says that any time a headline asks a question, the answer is NO.
And also QI!!
Seven of nine cats disapprove your comment.
10 out of 10 cats would like to walk on your keyboard. 9 out of 10 cats manage to. 4 out of 9 will hit "enter" when they do..
15 out of 10 will then knock your keyboard off your desk ... 😁
8 out of 10 people whose keyboard mysteriously ended up on the ground do not even *have* a cat...
O-o
(Yep, you're welcome.. lol)
A brilliant talk - thanks Jennifer and the RI!
Perfecty brilliant. Thankyou
Wonderful talk! Lots to think about. A huge thank you to Ri for sharing this.
Great lecture!
Absolutely fantastic
im gonna recommend this to the stats unit i tutor
Great content, great presenter!
This was really fun. She is a great speaker.
Great talk.
Good video. Thanks.
In this video you really made me understand much of the logic behind statistics, thank you for that.
What an entertaining talk! The examples she used are hilarious!
Very good video.
Good stuff!
There's a saying that if you beat a set of data long enough it'll admit to anything, and thats in pure science. In public domain where no one is familiar or have enough info to make an assessment, the snake oil sales men have a field day with them.
I seen it on the inside. When unfavorable results occur, when possible, they are discarded and called "outliers".
Real life example: When a traffic study showed a horribly backed up traffic pattern at a particular intersection, the study managers ordered the discarding of Friday results. "Fridays are always atypical."
Another factor is the number of surveys thrown away - if I throw away every survey that doesn't fit my message the stats aren't valid.
It's called P-Hacking
Curious about the null hypothesis. I certainly see Euler on the graph-y drawing, but it still smells like Gauss
EXCELLENT!!!
36:25 To be precise, whether the null hypothesis is within the confidence interval is a different thing than whether the outcome is statistically significant.
Bad statistics purveyed to the public much more often relate to improperly conducted statistical surveys rather than improper correlations drawn from survey results. In the social sciences and in the political sphere, improper statistical inferences typically are drawn if an insufficient sample of the target population is surveyed or if there is not a randomly drawn survey sample of the target population.
"How do we know it's the bacon"? Excellent.
@Anthony Butler The data are rubbish. Correlation does not imply causation.
@Anthony Butler Where are the controlled trials?
Nice one.
Reminds me of my math teacher who said...never trust a statistic you have fudged yourselve. 😂
A truly great lecture!
I wonder where I know her from. She seems familiar.
are you fan of Matt Parker (standupmaths)? She was featured in few of his videos.
Studies cost money, lots of money. So not many people are willing to pay for such a study unless they will get something that will help make a profit. So for any study or study results, one has to follow the money to find out what result was expected. That is likely the result that will be obtained... the old saying about drunks and lamp posts.
The number one tactic on misleading or dishonest studies is incorrect _assumptions._ They're never held to task.
Surveys are the biggest joke in life or science. There are expensive seminars on how to use them to provide cover for corporate actions, REGARDLESS of what the actions are. Deception via surveys is truly an art form which companies spend handsomely to exploit. The upshot is news reporting exactly like this : "A survey conducted by the esteemed Gallup organization indicated that the number one issue troubling Americans today is wife beating. 15 million responses were received, leaving a margin of error of under one-half of one percent." For this nonsensically exaggerated example but which illustrates the point, consider this the [fictional] entire survey:
"1. Do you approve of military actions to kill terrorists?
2. Do you approve of the space program?
3. Do you approve of wife beating?
That's all. Thank you for participating in our short survey."
All that has to be done now, is _just not be so obvious,_ and run something exactly along those lines of disingenuousness.
And it works every time.
Great! Few good laughs ☺️
Excellent stuff.
If we could get 50% of people doing this 75% of the time, Daily Mail circulation figures would plummet.
Stats is written to favour the client
Important and entertaining talk. A year and a half later and most COVID headlines show numbers per country not rates per thousand per country. Not enough headline writers have seen this video - unfortunately.
Just checked "How to Lie with Statistics" is copyright 1954 so explaining bad statistics has a long pedigree.
100% agree. Glad someone caught that... I thought I was the only one.
I took statistics for psychology in college instead of calculus and I will NEVER regret that decision. We went OUT into the world and took surveys, armed with a hypothesis and a clipboard. We could ask any question(s) we wanted. I asked: would you be willing to participate in a study? THAT was the study. If they said yes, I said thank you, carry on. If they said no (or ignored me) I said thank you for your input. EVERY response was a statistically relevant entry. This one question WAS the study. If someone stopped and asked to hear what the questions were, I told them. I wanted to see how many people would be willing to participate in a study. (Some were disappointed I didn't want to know more about their opinions) My hypothesis of 70% NO was verified with over 1500 queried, with about 1000 negative responses, 80% responded visually (shaking the head) or verbally, saying NO. The other 20% were comprised of people changing their path to avoid us, making a face at us, or using profanity towards us. I think those responses could be loosely interpreted as a NO. Disregarding the 3% of people who swear at people they like as a sign of friendship. My statistics professor told me IF you want to lie, use statistics, but if you want to prove something, use statistics. Also: Who is paying for this study? Who benefits from a certain result?
But you failed to gain the subjects informed consent (at least many of them), therefore you can't publish the results. lol
@@BuickDoc Consent is not needed in an anonymized survey. BUT, we were not out to publish, just get a good grade and learn about taking surveys. We got an A, but also best question voted by the class. The survey could only ask one question, how you got them to answer it was up to you. We needed at least 1500 responses and got around 1800 standing outside the shopping mall on 5th in SF. Also, this was in 1998. The question was an important one as well, indicating a resistance to surveys taken on the street, which were popular at the time. Soon after, phone surveys became much more popular. Lastly, it's not polite to laugh out loud at people you don't know, just because you think you know something they don't.
Fortunately we had an enlightened schoolteacher 60 years ago who taught us never to trust statistics, particularly those trotted out by politicians, advertisers, or anyone who a vested interest in influencing you.
and... did it help you to ignore the Covid hysteria of the last three years?
@@gordonspond8223 Although I accepted that Covid did exist, I was always sceptical of the weekly/monthly death toll figures regularly announced knowing that it included deaths from all other respiratory diseases.
Personally I took little or no precautions against having contact with other like minded souls and I kept a respectful distance from those who were clearly fearful of contact.
Good lecture but thank God I'm not so gullible as assessing the response of what's more dangerous, riding a bike or driving a car before knowing the basic facts:
1) A "bike", does that include a motorbike?
2) As well explained, measured by time or distance?
3) Only including accidents or bi-effects on health from the activity itself?
4) How is "danger" defined, only counting deaths or also traumas and even smaller wounds?
There are so many more factors needed to answer this question properly that any critical mind should reframe themselves to answer such a question based of that few statistics.
Later she is giving a fantastic lecture on the common misstake of mixing correlation and causality which is done over and over again by so called journalists specially when presenting themes of their own activism like climate change and crime statistics.
MCAS is what happened in 2018 with the Boeing 737 MAX (Maneuvering Characteristics Augmentation System), and has to do directly with the in flight trim system and was a software glitch that sadly cost hundreds of human lives but was traceable to human error, but I digress.
And I'd go so far as to say that the idea behind MCAS wasn't even bad at all, it was to address a very real potential problem with the changed configuration of the new revision of the 737.
The way it was handled was just terrible, not entirely sure how the type rating stuff goes in airplanes. But somewhere between the regulatory bodies and the manufacturer proper training on a new mechanism that was introduced should've been mandatory. Like at least one simulation session to experience what the system does and how to override if needed for some reason. Obviously this wasn't mandatory, I guess some airlines might've done it voluntary but that's not enough.
The implementation was also seriously flawed, I don't get it how the airline industry that typically has safety and reliability as a high priority could've implemented this only using one of the two angle of attack sensors present in the plane. It should've used both from the start and notify/disable itself when a discrepancy is detected between the two sensors.
@@extrastuff9463 There is still a lot of unknowns that won't be released until the results are finalized, and even though the changes were implemented over a month ago, the MAX remains grounded until they've received clearance from all aviation agencies and governments worldwide in order to ensure the safest and smoothest return to the air as possible, slated sometime in November last I heard.
One thing is for certain, after going through everything with a fine tooth comb, the Boeing 737-MAX will be one of the safest in Boeing's fleet and air travel still remains statistically the safest form of travel per person carried per mile.
I saw someone use a list of dog bite fatalities in the US.
The problem was that they did not show the proportion of dog ownership by species - so it may have turned out that the 'most dangerous dogs' were simply the most popular dogs.
I also think that getting that second dice roll because the first one did not match your hypothesis is well dodgy and you really should have been prepared for this and have had an explanation ready.
Slartibartfast she did: Her sample size was too small because the students stole the dice.
I scrolled through the comments a bit and maybe I missed something at the end. Didn't you say you would explain the hand shaking from the beginning at the end? I'm from Texas so forgive me if I missed it.
I think there should be more info on how she's characterizing things, because for example there is a shit ton more "alligators" and they're more widespread, but if you live in an area with hippos, it is likely that they are a greater threat.
@39:57 it's really quite a simple explanation. It's just captioned weirdly. And shouldn't have been in a pie-chart. But what they want to say is the following: in our survey today, 51% of the questioned people admit to have tried pot, whereas a year ago, only 43% admitted to have tried pot and in 1997, 34% admitted to have tried pot. If this has any statistically interesting values is another thing, as people would need to admit they did it. In 1997 it was probably frowned upon a bit more than now. So even if the same number of people actually did pot, the number would probably increase anyway. But that's a whole different kind of story. The statistic itself isn't that broken.
My Student pointed out that 25 cigarettes or more per day is being used to represent smoking ( see 7 minute mark). 2 rashers of bacon daily vs 25 cigarettes per day doesn't seem a fair comparison, does it? Also the logistics of smoking that much is amusing - assuming you have a whopping 12 hours of free time daily you'd need to be filling it with a cigarette at least every 30 minutes. I suppose such people exist, but I wonder. Otherwise, a good presentation.
Channels with videos interrupted by adverts have fewer views. Correlation = causation!
Nice working thanks ya.
6.2832 out of 10 statisticians agree: headlines and advertisements use clickbait statistical (mis)representations, just don't ask what happened to the 71.68% irrational statistician after participating in the survey.
Your numbers don't add up. They would if it were 2.832 that agreed; or maybe 9.2832 agreed and 7.168% didn't.
Time to buy a new calculator, Phroggster! 😛
@@jensraab2902 Nope, those numbers are fine... 10 parties surveyed, one party partially agreed roughly in accordance with the lesser-significant digits of τ, and was therefore deemed irrational.
@@Phroggster Like I said, your numbers don't add up. But if you keep insisting, then maybe the problem isn't your calculator after all.
@@jensraab2902 I understand that you didn't comprehend any of the joke layers presented, but 1-0.2832 = 0.7168 = 71.68%. As there's no other maths involved in my original statement, and you're deleting your commentary from the record, I bid you good day.
@@Phroggster Why should I be deleting my commentary?
Here's what you wrote: "6.2832 out of 10 statisticians agree"
The 0.2832 would correspond to 2.832 statisticians, not 6.2832.
I get the joke that there are no fractions of people (outside of crime scenes and morgues), that's why I didn't write "That doesn't make any sense." but rather "Your numbers don't add up." - and even specifically offering correct numbers.
But this is all getting really silly.
If you think your math is good, so be it. I'm outta here.
Crocodile vs Hippo.. how likely though are you to be injured by one when you come across one? I mean.. if you live in Florida, you may meet an alligator but probably never a hippo. Most of those alligators will just do their "I'm a tree trunk" thing. How many hippos, should you encounter one, would go after you? (Hint: they are notoriously territorial..)
I knew someone who worked at a safari park and he mentioned he'd rather be in with the big cats than with the giraffes...
You are not defining what you mean by "dangerous".
"Which animal is more likely to eat you?" Not a giraffe, and definitely not in Northern Norway.. a gator in the Glades might though.. Hippos are aggressive animals that are easily be provoked. I won't likely ever come across a wild one though..
Definitions of "dangerous"?
Not a lot of croc's in florida...lots of aligators...but crocodiles?
@@mickelodiansurname9578 Okay true. XD My spouse lived there, I'm glad to avoid the bugs and snakes and Gators though lol
There is an error at 5:32. It's "400 random individuals" not "400 individuals who didn't eat bacon".
Airlines assign seats by computer to load balance and optimize lift and fuel economy.
Les Prigmore the point is that Ryanair assigns only the “middle” seats by computer to load balance and such, leaving the rest open to maximize profit. It wouldn’t be so bad if they were just honest. In fact, it might even be more attractive to pay the extra for the guarantee of a better seat.
About the ryanair, there is nothing wrong. They never say "random allocation". They say you get a seat. They never really lied. I dont like them, but they didnt lie. It is customers assuming.
The increase from 0.6% to 0.7% is 0.1 percentage point, but almost 17 percent. I liked the lecture overall, but statisticians should be very careful not to confuse the two.
She talks precisely about that right at the start. It's a 16.67% _relative_ increase, but only a 0.1% _absolute_ increase.
There are lies, dammed lies and statistics.
That is what I was taught in my first class of statistics.
Peter Scheen the quote is there are lies, damned lies and government statistics.
That quote was from Mark Twain
@@neildahlgaard-sigsworth3819 Not originally. According to Wikipedia, the earliest known usage was: "The Derby Mercury (Derby, England), October 21, 1891; Issue 9223 "Sir Charles Dilke and the Bishops" "A mass meeting of the slate quarry-men of Festiniog [Ffestiniog, Wales] was held Wednesday night [Oct. 14] to protest against certain dismissals from one of the quarries...." He [Dilke] observed that the speeches of the Bishops on the disestablishment question reminded him that there were three degrees of untruth--a fib, a lie, and statistics (Laughter)"" en.wikipedia.org/wiki/Lies,_damned_lies,_and_statistics
Apparently it was a variation on an earlier phrase, "Liars, damned liars, and experts."
@@neildahlgaard-sigsworth3819 No, not government statistics, if you are really interested I will show you some statistics from anti vaxers where they misrepresent the results of vaccinations.
@@stephenlitten1789 I know. although I was told it was Oscar Wilde.
39:57 , it's pretty simple to understand, even if the pie chart is still wrong. They meant: "Americans who (declare to) have ever tried marijuana are 51% today, with respect to 43% last year, and 34% ten years or so ago (1997)". There was no reason to put those data in a pie chart, of course.
Cycling versus driving - define “danger”. Of injury? Of death? Of injury or death to others?
52% of 52 people?
so 27,04 people?
27/52 = 51.9, so the company rounded up.
@@MN-sc9qs Right. It's either that or they have to report that "51.92307% of people..." etc.
Some of these variations you reference should be codified into the list of fallacies argument.
I'm not from the U.K. so I'm just curious: can someone tell me what part of the country her accent is from?
Manchester?
I get a mix of Manchester (Brian Cox) and Geordie (Sarah Millican)
Eat bacon, but never eat cheese before you go to bed 😄❤️
But muh night cheese...
It ain't easy being cheesy ...XD
I enjoy nightmares.
Talc powder and Ovarian Cancer-no controls; And don't get me started on the COVID distortions.
There's a 100% chance that I will continue to eat bacon sandwiches now and again!
The bacon survey would have been even more interesting if it included whether or not brown sauce was on it or not (where possible it should be).
*Bacon/Brown sauce joke removed by author* not appropriate for RI
Forget the correlation. Are you telling me in 2008, around 800 US people died by being tangled in their bedsheets!! What!! How do you even do that!?!?
Anyone else see IT (Pennywise) on her top!?
Also. Good to hear an accent for a change; yorkshire if I'm not mistaken!
❤❤❤
I ride a motorcycle every day, work with chemicals for a living, and skydive for fun. Your statistics can suck it. :)
@bagariddum They're minimized, just like looking both ways before stepping into a street. I'm much less likely to be harmed than the average person that's always checking their phone while driving.
The standard reply to this sort of thing tends to be "the plural of anecdote is not data". Wish you best of luck nonetheless of course...
@@AttilaAsztalos Um... Yes it is. That's exactly what it is. Or are you making a clever jest that I missed? Sarcasm is hard to pick up in type.
@@AttilaAsztalos or survivor bias
8:15 did they really worked with people who eat bacon every day? Actually, how do you measure that? Say you have subjects who eat bacon every day, do they follow up with their lives until their deaths and count how many people developed cancer? Obviously not. Whatever the methodology in the research is, is it fair to interpret the results as prof Rogers did? I don't know, I would just really like to see it addressed more closely.
They ask a bunch of people how often they eat bacon and a bunch of cancer patients how often they eat bacon and then compare the numbers.
41:42 Ben and Jerry's... consider the source...
I think to start a business in Bed sheet Insurances for those who like to eat cheese before going to bed!
Great show but for one or two things. The very first thing she did - the crocodile v hippo thing - was done improperly. She first asked which animal is most dangerous - which is a function of the individual animal, i.e. how likely is it that you will be injured if you encounter (1) a crocodile and (2) a hippo? She then said which one causes the most deaths (in humans is apparently what she meant)? That number, 1,000 v 500, is a function of the human, not the animal. Doesn't matter if there are 1 million crocodiles but only 100k hippos in the world - there will still be a 1,000:500 (or 2:1) ratio of human kills per year - sure sounds like the croc is more dangerous - but, in fact, it is 5 times LESS likely to kill a human (using my numbers). The likelihood of being killed in a single encounter shows the hippo as the far more dangerous animal: 1,000:1m v 500:100k or 1:1,000 v 5:1,000. The hippo is the more dangerous animal. The moral: anyone, including this presenter, can manipulate the stats to prove a point whether that statement is factually correct or not.
Moe importantly - the crocodile number refers to *world-wide* "crocodilian" incidents, including alligators and caimans. Since Hippos only live in Africa while crocodiles, caimans, and alligators can be found in Africa, the Americas, Australia, South-East Asia and India, the numbers are heavily biased.
It's like comparing the risk of getting maimed by a land-mine (anywhere) to the risk of getting maimed by a circular saw in Britain. The land-mine risk would be significantly higher, even though there are no land mines lying around in Britain...
I'm surprised she didn't mention the obvious attempt to mislead people that was "52% of 52 people" (they're hoping that people will read it as "52 of 52 people").
52% of 52, it was meant to make people believe all 52 people agreed
Is that a Manchester accent?
I don't quite have ALL the regional dialects down yet, so it'd help if someone could definitively answer yes or no to that.
Razar Campbell more like Hull maybe?
@@stilltherealvicz
Cheers, mate :)
I'm curious too, it sounds somewhere between Sarah Millican (Geordie) and Brian Cox (Manchester).
My take away from this is ... Ben & Jerry need to make bacon flavored ice cream !
In your claim that London isn’t as violent as New York you say that you need to look ‘over a larger period of time’- but that runs counter to the claim that a London Has Become more violent. Looking over a longer period of time is irrelevant since we’re not asking which city- through the course of its existence- has had more violence. We are looking at local trends and their shifts in the here and now.
If you extend the time period to 100 years you're about 70x more likely to be killed by a military/civilian destructive act in London than in New York. About 3x more likely to die in a plane crash in New York than London.
This all ignores the fact that New York is about 250x the size of London; London's smaller than most the Boroughs of New York.
The important thing to remember here is that she probably means "on average". On average it's safer to live in London than in New York. That's why she says to look at a larger period of time. It's nonsensical to look at local trends. She even tried to solidify it later with the brief explanation of 95% confidence intervals. The fluctuations might just be random chance.
I'm so glad I was born with the 'Common Sense Gene' !!
To my surprise there are at least as many hippo's in the world as there are crocodiles.(if there were twice as much they would not - per individual - be more dangerous rhan hippo's
RE flying. Aeroflot is one of the safest airlines statistically.
the crocs vs hippos things needs a per encounter or similar qualifier though like the others have - otherwise it's utterly meaningless and a case of bad statistics, which this talk is supposed to be against
52% of 52 people agreed to the twice daily thing!
"every time you have a CT scan it doubles your risk of xx"
me: "not really true that it doubles or triples, it does cause significant damage to cells but the cells are able to cope with it for the most part, however the mystery of not knowing where the brain bleed is and how bad it is will kill your child faster than any cancer would, guaranteed, operating on anyone increases the risk of spreading infections from surface to internal organs, it increases the risk of surgeons getting sick from patients that are infected, it's not exactly a safe world but the medical world acknowledges these risks and have deemed them worth taking the risk, especially when there's no alternative.
Aren't you glad there's ultrasound to find out how developed your fetus is hmm? would be much safer than the 3rd alternative of opening up the mother, the second is waiting till birth.
The fact is that if there's a less risky alternative to a situation that is less problematic then those will be given first, such as keyhole surgery to mend broken bones or heart surgery vs a nice slice into your body causing the infection rate to go up 10,000x
Your risk of dying tomorrow is the same as it was yesterday, live your lives ffs or else there would be no point to having a heart beat.
and what I have said has not only been studied on, but observed repeatedly by medical professionals both sceptical and open minded and agreed to the same degree
52% of 52 people = 27.04.
How on earth can you have 0.04 of a person?
Also 74% of 54 men = 39.96 men.
Why are you not even questioning the very basis of these numbers?
Because of Rounding
It's not machining, decimal place is irrelevant.
When it comes to population sampling, it's about *representation*, and probability of that representation.
Nothing to question about, you just don't understand the applications.
So what that means is when you read 52%, meaning one group is majority, Which is either 27 or 28 people from that sample. Nothing more!
Or is it! So if you eat ice cream wile swimming an get brain freeze.... stadiums have lots of stairs...sticky Cheese? LOL This is a good video! I am silly in this comment but now I will regress to the norm. Oh, wait I am not normal.
Hippos used to get credit for 4000 deaths per year
there are crocs in Australia, but no hippos
@@balthazarbeutelwolf9097 Oh wow! That's a great point. I can't tell you how many times I've been confronted with the "assumption" busting "statistic" that hippos kill many more than crocs. But I believe it was a statistic that related to Africa, NOT the world. Excluding Australia of course.
@@pheresy1367The crocodile deaths are *definitely* world-wide and also include *alligators* and *caimans* ("crocodilians") thus including Australia, India, and the Americas. So the statistic is heavily skewed against the crocodile.
Here's the source by the way: www.crocodile-attack.info/about/human-crocodile-conflict
It's a bit of a misleading example. Cycling is more dangerous than driving a car MAINLY BECAUSE of car drivers ;-)))
Telum Atramenti It’s not misleading. It’s true. In the current road environment many of us live in cycling is more dangerous. Cars also have to contend with other cars on the road. She wasn’t showing a comparison of the danger of cycling vs. driving when there are no other vehicles on the roads.
it doesn't matter. If you die, you die, regardless of whether you were bumped by a car or fell in a ditch.
@@Harry351ify How many brackets does it take before you people understand when a remark is meant as humour? 500? Or would 15 be enough? )))))))))))))))
@@Harry351ify Not to mention that no, it's not strictly true either. There are situations in which it actually does matter. Whether an activity itself is INHERENTLY dangerous or the circumstances in which you engage in the activity MAKE IT dangerous. Driving is INHERENTLY dangerous activity unlike biking. When you're inside a vehicle which can accelerate to over 120 km/h in mere seconds, and which weighs over a ton - the activity of driving becomes a very risky business regardless of where you drive, - country road, residential area or highway. Lose control, mistake gas for brake pedal at the wrong time and place, etc. and you're done for. Not so with the bike which a slow moving and very slowly accelerating vehicle. When you are biking for leisure and are using bicycle paths in the park - you remove up to 90% of danger associated with the activity of biking. In fact the whole dilemma I am describing is EXACTLY what her lecture is about: which is how the stats are to be INTERPRETED.
In my city, the greenest city in North America - there are many parks with bicycle paths crisscrossing the entire city North to South and East to West , there are also a number of bicycle lanes separated by concrete embankments from the rest of the traffic. If you favour a FACE VALUE interpretation, - you will take a car to work, thinking it's safer and you will most likely make a bad decision. Instead, you could purchase a bicycle and take the park paths and embankment-protected bicycle routes to work, and you will in fact be not only safer but healthier and will probably live longer too. It's exactly the same idea with interpreting stats about cooking being a fire hazard without taking induction stoves into account, or with deaths attributed to opiates which did not take ROA, criminal law, Fentanyl adulteration, mental illness, suicide and other circumstances into account, and the Procrustean interpretation in the end hurts patients who require pain control. Not only can there be bad, overly generalised interpretations of data, but also Procrustean, badly put together studies which do not allow to isolate the cause of the problem at hand.
So a ham sandwich will be OK for my lunch tomorrow ?
Probably
But if it has cheese, strip the bed of sheets.
20:35
Ole's At tHe wHEeL
... but the only times that aren't "before I go to bed" are when I'm actually in bed.
Personal favourite: "air travel is the safest form of travel" measured by crashes per mile traveled.( ie, - a flight of 10,000km which ends in a fireball on landing is a 99.99% safe flight.)
That card about hippos and crocodiles is BS. First, the number given for crocodiles includes other animals (like alligators). Second, it doesn't take into account _exposure_ to those animals. There are no hippos in Australia, for example, so over there the "risk" is zero. In places where both species exist (parts of Africa), hippos actually cause a lot more (human) deaths. Not a very good way to start a lecture...
This is why I don't read newspapers.
Lol.
This is why you are dumber than the average - because of understanding how news work, and what they bring to people's understanding of reality, and to their freedom, you renounce it completely, which makes you inside your own bubble, which is way more difficult to break by yourself than if you were reading different newspapers every day.
Which eventually makes you ill-informed compared to the average, while believing you're better informed.
@@nicojar
Lots of assumptions there.
I didn't say I wasn't informed... just that I have never read newspapers.
I pretty much stopped reading newspapers 20 years ago, because broadband Internet.
My goodness she's beautiful ❤️ really interesting presentation
I was literally looking for a comment like this. Very beautiful woman, fantastic intelligence, great accent. I call a woman like that a catch.
Having 1 extra person out of 400 dying of pancreatic cancer just because they chose to eat bacon doesn't seem like much, but if you consider the EU population of 500 million citizens, it means losing an extra 1.25 million people a year because of poor dietary choices. Again, it is easy to manipulate numbers if you don't look at the big picture. Please eat more fruits and vegetables, nuts and pulses, for your own sake.
Accurate analysis of the situation may make the mathematical results more or less significant.., there's the real problem.
advertisers are selling a dream, illusion is King and your deflationary
Surveys are the biggest joke in life or science. There are expensive seminars on how to use them to provide cover for corporate actions, REGARDLESS of what the actions are. Deception via surveys is truly an art form which companies spend handsomely to exploit. The upshot is news reporting exactly like this : "A survey conducted by the esteemed Gallup organization indicated that the number one issue troubling Americans today is wife beating. 15 million responses were received, leaving a margin of error of under one-half of one percent." For this nonsensically exaggerated example but which illustrates the point, consider this the [fictional] entire survey:
"1. Do you approve of military actions to kill terrorists?
2. Do you approve of the space program?
3. Do you approve of wife beating?
That's all. Thank you for participating in our short survey."
All that has to be done now, is _just not be so obvious,_ and run something exactly along those lines of disingenuousness.
And it works every time.
Now you know.
And, um, if it matters, GOVERNMENT actors do this to persuade a bill to pass.
Especially if there is funding. Or jobs (==votes).
Believe that or not. Who's being naive, Kate?
Cancer is only one of the risks..
'52% of 52 people agreed' - the marketing bulsh!tters wanted audience to see 52 people out of 52 people agreed (i.e. 100%) hoping they wouldn't notice the % mark. Devils.
There is a 99.999% chance that there is no climate emergency which will wipe out humans in the next 5,000 years 🙃
Just to be pedantic, you spelt disagree wrong in the presentation.