That audience was so non responsive. They didn't raise their hands or laugh at her jokes. Were they all sleepy after eating a big lunch and ready for a nap?
I think she tried too hard to be funny. While beeing funny is good, it should just be used as a method to transfer information instead of substitute it. What did here fart joke contribute to the topic she was talkign about? Why did she gave the 5 "lies" in statistics cool sounding names, instead of names that actually refer to the problem? Putting two graphes with different scales on each other to put the relations in a misleading proportion to each other is called "the good looking graph". You can use this technique also to make a graph look shitty and the name does say you nothing about how this mechanic is applied or what to look for. There is no statistical self defence in here. So while her topic was really interesting and she had good points in it (like 6:55), but in general the presentation seemed often more forced than informative, even if well-meant.
yosoy1loco Because most people there are not of her specialty, so providing a dry lecture is a good way for people to gloss over your presentation. Few people like to be presented with a vomiting of facts like traditional textbooks do. So you have to use rhetorical devices to maintain attention. Humor is such a device.
Necrostrike I completely agree^^, but humour cant inherently be forced - that was my argument. I think she would have been funier if she tried less to be funny. Wasnt meant offensive,
Wow. She explains how to spot estadistica fallacies in the news. And at the end she pull one of those on the attendees. And nobody even noticed. She used the exposure of a stadistical fallacy to validate another one.... Low
She spoke a lot about how you grab information at what we can see or the headline without second-guessing. I learned that most people will alter the way they present their data by their own personal stance or benefit. I am someone who will read something and automatically believe it because hello people should be giving out the correct information. But over the past 8 years, I have had to really stop taking people's word at face value. I learned in this video to double-check where they got their information, their source, or what they put in the questionnaire is more important than the title they posted to get views.
Beware of numbers versus percentages. Local hospital was highly criticised for going several million over budget. Turns out, compared to the annual budget, the overspend was the equivalent of less than ten hours spending.
Yeah so true, it drives me mad in science like in my field astrophysics you'll hear every year the same endless debate about how nasa is overfunded more then 20 billion we need to save money blabla. Even if i grant them the fact that the funds they are diverting from scientifoc reaserch will be better used elsewhere -they won't- we are arguing over a few hundred million, a few billion at most which is so small if you consider the us. Nasa costs less then simply maintaining our nukes, those cuts proposed that would have haltered james webb project and scientific progress with it were less then what the army spends on bullets, the nasa budget is probably a weeks worth of us military spending but when you talk big numbers you're sure to arouse idignation and panic in people
Numbers are either relative (1 out of 10; 95% percentile) or absolute (5 units; 1,336,670 elephants). It's easier to evaluate numbers when they're relative; otherwise, if you're looking at an "absolute" number, my advice is "no lonely numbers." (I stole that from Hans Rosling.) Otherwise, it's so easy to manipulate and make that number appear as large or small as you want: it all depends on what it's being compared to.
I took statistics many years ago and also read a book on how statistics can deceive. I found the subject to be more interesting than the usual dry math course.
Pretty simplistic presentation without a lot of data. Basically don't trust correlations verse causation, unless it supports your particular view. The author was guilty of the same thing she was supposedly warning about with her antidotal alternative interpretations of the data. Geez! Does anybody know how to think anymore?
Wow. She explains how to spot estadistica fallacies in the news. And at the end she pull one of those on the attendees. And nobody even noticed. She used the exposure of a stadistical fallacy to validate another one.... Low
Sorry dear internet, but the Dutch are terribly uncharismatic and bland. Our cheese makes up for it somewhat, but I wouldn't try to learn presentation skills here - even our best don't know what they're doing.
I noticed something. She presented two data sets about drug offenses/use grouped by ethnicity. Let's call the data sets A and B. She then argued that since we have data set B which disagrees with data set A, we should really question the immediate surface interpretation of A. I can't think of a reason not to do it in the other direction as well: given that we have data set A, we should really question data set B as well. Unless of course exactly one of the data sets agrees with our preconceived notions, in which case we should just dismiss the data we don't like :P
It's more about seeking and analyzing the right data set to answer the question at hand and making sure that they are representative of the population you wish to study. If you'll allow me to elaborate: Let's say we wan't to answer the question of whether there is a difference in the rate of homelessness when comparing ethnicity. The data being used to answer this question is a federal-wide census that aggregates the individual state census data collected over the year by each state. Let's say also that the statistician responsible has ignored the proportion of each ethnicity within the population of the state and has only reported an absolute number of homeless persons by ethnicity (e.g. State X Census: Caucasian 45k Homeless, Latin American 33k Homeless, African American 46k homeless, etc.). Once this data is aggregated, the same absolute results are used to represent the entire population of the US. This leads to any statistics derived from this data set to inaccurately represent the *proportion* of homeless people by ethnicity (i.e. adjusted for the size of their individual strata). Since only the absolute numbers are being observed, this can also lead you to make false claims about the populations you are studying. This is called sampling bias or sampling error.
Yeah, it's not like there could be some kind of "bias" affecting how police deal with a group of people who earned equal rights status a few decades ago. Inconceivable.
Wow, she was doing so well and then bombed at the end. She actually took a survey of personal responses for something illegal and used it to try and prove a point, without providing any context. She also forgets most of these arrests and charges are usually for or combined with other offenses and can get pleaded down to a single charge and many other circumstances including who is interpreting and writing the paperwork; Simpsons paradox anyone...
She oversimplified the facts behind the troubling statistic at the end, perhaps to cater to the people who believe the justice system is skewed against minorities. It may or may not be but there is more to it.
Wow. She explains how to spot estadistica fallacies in the news. And at the end she pull one of those on the attendees. And nobody even noticed it. She used the exposure of a stadistical fallacy to validate another one.... Low
She is amazing, such a professional in spite of the feedback she's receiving. Perhaps the quiz to see if anyone knew what Technology, Entertainment, Design meant rubbed them the wrong way🙄
Poverty leads to crime, groups that have higher rates of poverty will have higher rates crime. It is less about racist police and more about economic disparity. That is to say the issue is not that the stats are over-inflated per se, it is that they are proportionate to rates of poverty which leads people to criminal activity and rates of poverty are not equal among all groups.
That's exactly what she was talking about. That's a good hypothesis you have there, but do you have any data to back it up? It's easy to come to conclusions and even believe them to be true, but are they?
I'm currently ending my 4th semester at political science (where we also learn statistics) and it's HAAAAAAAAAAAAAAAAAAAAAAAAAARD as @$%# to prove that anything has causality. Mostly the goal is just to reduce the chance of there not being causality by a lot and then assume that there's causality. And this is what even PHDS do and our leading scientists. They're just so hardcore that they can reduce the chance of failure by very much but they still can't guarantee causality.
@@drewpocernich2540 If it can't be denied, then you can't causally infer the link between political science and what defines something as a corrupt man's philosophy :p
Just to be clear at 5:06, there is a 12% increase in abortions and a 52% decrease in cancer screenings from 2006 to 2013 according to that set of data. Plotting things on the same scale is fairly irrelevant. I would hardly say 12% increase is "unchanged."
Her presentation was excellent! I don’t know why the audience was so dry, but I wish she would have the chance to do her presentation at another TED location. Outstanding job on her part.
Ok til the end when she felt it necessary to editorialize about a huge topic, law enforcement policy and behavior, had to drop it from my recommend list. Glad to see others here in the comments section saw the same thing - over-exuberant social justice warrior - lost cred.
Anyone else waiting to hear the "results" of the poll she did in the beginning? Was expecting her to say something like "only 80% of TED Talk listeners have ever heard about TED" or something, since a lot of people for one reason or another didn't hold up their hand.
Several times in this boring and unnecesary talk, she points out somebody lying with statistics, and then posits an equally unsupported explanation rather than just leave it at there is no evidence for that.
Her representation of the processed meat increasing bowel cancer risk by 20% is also a misrepresentation. One factor she doesn't consider: population number. So while your personal absolute risk goes for 4.4% risk of bowel cancer to 5.2% risk of bowel cancer if you eat 50g of processed meat per day, over a population of over 300 million that .8% points is absolutely massive. 300m x 0.8% = 2.4m people
@@HHumii if each individual's odds of developing the condition goes up by 0.8%, you would expect to see 0.8% more people develop the conditions with a large enough sample size. If you don't see why intuitively (skip this paragraph if you do), consider that we can determine the odds of something happening by taking the number of times it did happen divided by the number of times it could have happened. For example, if you were to roll a dice 600 times, about 100 of those rolls will be 1 - you can determine that you have a 1/6 chance of rolling a 1. The reverse is true as well - if you know you have a 1/6 chance of rolling a 1, you can expect roughly 100 of any 600 rolls of a dice to be 1. This applies here, too. If the odds of developing cancer in each individual increases by 0.8%, you would expect to see roughly 0.8% more of the population develop cancer. In a large population, if the total number of people developing bowel cancer increases by 0.8%, though it sounds small, can be a very large number of individuals effected. 0.8% of the US population is around 2.4 million people - that's 2.4 million people who otherwise wouldn't have gotten bowel cancer. That 0.8% can be massive on the scale of an entire country.
@@HHumii It's relevant on the scale of the whole population, which is usually what these statistics are about. I think it's really weird how she dismisses a 20% increase (that actually IS a 20% increase!) because the numbers "only" go up from 4.4 to 5.2% - that's a HUGE effect on a national scale (let alone an international scale) and honestly, it's not irrelevant on a personal scale either. What she didn't mention is that, the incidence of other types of cancer and cardiovascular disease go up with the consumption of red/processed meat, so yeah, cutting out processed meat, can make a huge difference for an individual as well, over the course of their lifetime.
Uh that was uncomfortable watching it I could only imagine being there, she really should have showed the results of the 3 questions she ask the audience before hitting those high level touchy subjects
1.Statistics are more involved in making choices in our daily lives than we think. Therefore, we must look at the statistics carefully 2. Because statistical graphs are easy to manipulate, the source of these figures should be investigated. 3. Statistics can change over time because a lot of information is gathered and extracted 4. When interpreting statistical data, consider whether there were other events at the time of the event for that subject. 5.Misinterpretation of statistics can hurt others. Therefore, we should look at the statistics from a critical perspective.
I was on holiday in the USA and was waiting for my wife outside a shopping mall. A young lady approached me and asked if I would mind answering a few questions on the products for sale. My answer was 'No thanks' I'm a tourist here and I have no experience of using American products. Her reply was 'That's quite ok, but you get 6 Dollars for completing the questionnaire and it takes just 4 minutes to answer. I then asked her if she was earning commission on every completed questionnaire to which she replied 'Yes of course' Is it really any wonder that just hearing the word 'Statistics' is interpreted as an exercise in deceit and dishonesty to anyone who thinks for themself.
The last explanation of the drug offenses doesn't really make sense. Taking drugs is just a small portion of "drug related offenses" (selling, delivering, consuming). But since her explation is just about the consumption, it doesn't prove any racism.
Judea Pearl’s “The Book of Why” would be an excellent basis for a short course in defending oneself against misinterpretations of statistics, and even better how to better interpret them.
Wow. She explains how to spot estadistica fallacies in the news. And at the end she pull one of those on the attendees. And nobody even noticed it. She used the exposure of a stadistical fallacy to validate another one.... Low
Something is wrong and don't blame the audience! Sanne is trying to be natural, but it's all calculated and she's unnatural. Something in the atmosphere sucks. It's a fail! Sure she got some knowledge and it's an interesting subject, but being politically correct and nice in order to have her head patted is just lame (ref. chapter 2 of her book).
Thank you. Very good introductory talk help us better see and arm ourselves with critical thinking skills to help defend us against the damned lies held in some media driven statistical data. Thought the ending was good too and needed saying. Brave woman.
im glad she brought up the integrity of the data as well, because even with these statistical differences there could be bad data to start with like including anything near school with a loud bang noise as mass shooting lol. many others of course but you get the idea media is really good of making things ten times more crazy
"Newer - Betta- DATA"... I love way she points out that not ALL things come from an opinionated reporters "hard data"... you can eventually prove any of your points , just find the right data..LOL
How to prove a point about the fallacy of using statistics without context to then use a statistic without context to disprove another statistic also without context while also drawing conclusions.
Now we need a video titled: "How to defend yourself against misleading titles on UA-cam". When you say "How to..." there should be an actionable task to follow.
Her #5 is flawed. Correlation ≠ Causality? If this were always true, then there'd be no such thing as science. It's more accurate to say that correlation does not guarantee causation.
"Correlation IS DIFFERENT from causality" does not mean that there are no causal correlations. Is like saying "swans =/= white", it does not deny even that there's a majority of white swans.
@@petitio_principii And you just illustrated a perfect false correlation between your swan example and the original argument. Every effect in science is an instance where correlation = causation. This is very important to understand. But when you have virtually no correlation (except the accidental variety), you can have NO causation. Example: a rise in UFO sightings coincided with a sharp rise in the stock market. This is an accidental correlation, not a cause-and-effect correlation. But when you pour two chemicals together and get a pink liquid, every time you pour the same two chemicals together, that's a cause-and-effect correlation. It's how the universe works! Hope this helps.
@@RodMartinJr correlation itself never "equals" causation; it can only be _"caused by a causal relationship."_ To say they're the same is simply a categorical mistake. Correlations are only "clues" to eventual causal relationships, you can have causal relations masked by other factors that will result in an absence of visible correlation or even a _negative_ correlation between cause and effect. When you can filter those factors out, the correlation will be seen, but it is still true that correlation is one thing (co-incidence) and causality is another (the actual mechanics that should produce non-random patterns, correlating cause and effect).
@@petitio_principii Well said. I agree. You're using more precise language, and I love it. But I stand on my first statement. Too many people disconnect correlation and causation, missing the most important connections that are the foundation of science. Take the "Climate Change" scam, for instance. Sure, there can be other things masking the CO2 effects on warming. I don't doubt that for a minute. But CO2 is pretty wimpy for its "signal" to be invisible on virtually every time scale but one -- and in that one, temperature is driving CO2 into and out of the oceans. The warming of the last 300 years and the increases in CO2 are largely, if not only, an accidental coincidence. The Modern Warm Period is the *_Coldest_* of the Holocene's 10 major warm periods (1,000-year cycle), and CO2 couldn't keep the Modern out of *_Last Place!_*
That is absolutely true, evenmore so in some cases a correlation of zero guarentees independence etc. However the problem is that in the real world the statistics requires one to make assumptions and infer a distribution which can be incredibly hard. Most journalists and whatnot have no clue how immensely complex statistics actually is
I noticed something-again. At 14:40 she presents drug arrest numbers for whites and blacks: 332 vs. 879 (out of every 100,000). At 15:23 she presents illicit drug use statistics: 9.5% for whites, 10.5% for blacks, a difference of 1%. Blacks use more drugs than whites according to the data she presents. Then we should expect the arrest rate to be higher for blacks than whites. Now, 1% of 100,000 is one thousand. If every drug user was arrested, should we then expect numbers of 9,500 for whites vs. 10,500 for blacks? I don't think so-if the arrest figures are yearly, if every illicit drug user is immediately arrested and given a two year prison sentence (and they don't escape), we should see those large arrest figure in one year and 0 in the next (for both racial categories), or maybe you'll have one half of druggies in prison and the other half roaming free, with the two groups switching roles every year, so you have stable arrest figures of 4,750 for whites and 5,250 for blacks. And in reality the arrest rate is less than 100%, and so on and so on, there are many complicating factors. But if the drug use rate is higher for blacks, I would expect the arrest rate to be higher for blacks. Maybe the arrest rates should be proportional to the illicit drug use rate; 10.5% is higher than 9.5% by about 1%, which is approximately 10% of 9.5%. So the arrest rate for blacks should *maybe* be 332+33.2 = 365.2 (per 100,000), and we can argue about the reason for the gap between 365.2 and 879. Maybe 365.2 is wrong for subtle reasons. But this stuff should be argued in considerable detail before you jump the gun and conclude "MUH RAAAAYSISUM". She claims the drug use is the same for the two ethnic groups. I assume this is because the study giving us the 9.5% and 10.% numbers show a difference which is not statistically significant. In which case my analysis is somewhat moot. But then she should explain why she displays two different numbers yet says there's no difference. Also, how to lie with statistics #637: present two sets of data using different units, such that the absolute numbers are very large in one case (332 and 879) and very small in another (9.5 and 10.5).
Yet she jumped to an unsubstantiated conclusion by calling police racists. Perhaps there is another answer, but she didn't want to pull the data to check. She turned a decent talk into a thumbs down with her unwarranted political commentary at the end.
Who else thinks that the audience was very unresponsive and she deserved better?
Great speech. She just told the truth and the audience was disappointed 'couse peoples prefer false certainties vs. uncertain truths.
5 lies:
1. 3:51
2. 5:24
3. 7:26
4. 10:00
5. 12:10
Good presentation. She also tried to blend some humor there. Alas! Audience were never engaged.
Looked like half the audience didn't raise their hand on the TED question lol
That audience was so non responsive. They didn't raise their hands or laugh at her jokes. Were they all sleepy after eating a big lunch and ready for a nap?
What a bad audience, this was a great presentation and they were so dry.
Necrostrike well that's Dutch people for ya
I think she tried too hard to be funny. While beeing funny is good, it should just be used as a method to transfer information instead of substitute it. What did here fart joke contribute to the topic she was talkign about? Why did she gave the 5 "lies" in statistics cool sounding names, instead of names that actually refer to the problem? Putting two graphes with different scales on each other to put the relations in a misleading proportion to each other is called "the good looking graph". You can use this technique also to make a graph look shitty and the name does say you nothing about how this mechanic is applied or what to look for. There is no statistical self defence in here. So while her topic was really interesting and she had good points in it (like 6:55), but in general the presentation seemed often more forced than informative, even if well-meant.
Luckily they still clapped when she was done. But, yea. The Dutch people aren't that enthusiastic :p
yosoy1loco Because most people there are not of her specialty, so providing a dry lecture is a good way for people to gloss over your presentation.
Few people like to be presented with a vomiting of facts like traditional textbooks do. So you have to use rhetorical devices to maintain attention. Humor is such a device.
Necrostrike
I completely agree^^, but humour cant inherently be forced - that was my argument. I think she would have been funier if she tried less to be funny. Wasnt meant offensive,
My high school civics teacher said, "Figures don't lie but liars do figure".
Wow. She explains how to spot estadistica fallacies in the news. And at the end she pull one of those on the attendees. And nobody even noticed. She used the exposure of a stadistical fallacy to validate another one.... Low
She spoke a lot about how you grab information at what we can see or the headline without second-guessing. I learned that most people will alter the way they present their data by their own personal stance or benefit. I am someone who will read something and automatically believe it because hello people should be giving out the correct information. But over the past 8 years, I have had to really stop taking people's word at face value. I learned in this video to double-check where they got their information, their source, or what they put in the questionnaire is more important than the title they posted to get views.
Beware of numbers versus percentages. Local hospital was highly criticised for going several million over budget. Turns out, compared to the annual budget, the overspend was the equivalent of less than ten hours spending.
Yeah so true, it drives me mad in science like in my field astrophysics you'll hear every year the same endless debate about how nasa is overfunded more then 20 billion we need to save money blabla. Even if i grant them the fact that the funds they are diverting from scientifoc reaserch will be better used elsewhere -they won't- we are arguing over a few hundred million, a few billion at most which is so small if you consider the us. Nasa costs less then simply maintaining our nukes, those cuts proposed that would have haltered james webb project and scientific progress with it were less then what the army spends on bullets, the nasa budget is probably a weeks worth of us military spending but when you talk big numbers you're sure to arouse idignation and panic in people
Numbers are either relative (1 out of 10; 95% percentile) or absolute (5 units; 1,336,670 elephants). It's easier to evaluate numbers when they're relative; otherwise, if you're looking at an "absolute" number, my advice is "no lonely numbers." (I stole that from Hans Rosling.) Otherwise, it's so easy to manipulate and make that number appear as large or small as you want: it all depends on what it's being compared to.
It's always important to look at how the data was collected.
This is why critical thinking is important.
You are only demonstrating how weak and easily manipulated stats are.
I took statistics many years ago and also read a book on how statistics can deceive. I found the subject to be more interesting than the usual dry math course.
Can you name the name of the book? I am really interested in the matter.
"Damned Lies and Statistics"
Pretty simplistic presentation without a lot of data. Basically don't trust correlations verse causation, unless it supports your particular view. The author was guilty of the same thing she was supposedly warning about with her antidotal alternative interpretations of the data. Geez! Does anybody know how to think anymore?
Mark Twain: ther are three great lies - lies, damn lies, and statistics.
And he attributed it to others.
Don't get me wrong, I love this quote!
Wow. She explains how to spot estadistica fallacies in the news. And at the end she pull one of those on the attendees. And nobody even noticed. She used the exposure of a stadistical fallacy to validate another one.... Low
Sorry dear internet, but the Dutch are terribly uncharismatic and bland. Our cheese makes up for it somewhat, but I wouldn't try to learn presentation skills here - even our best don't know what they're doing.
I noticed something. She presented two data sets about drug offenses/use grouped by ethnicity. Let's call the data sets A and B. She then argued that since we have data set B which disagrees with data set A, we should really question the immediate surface interpretation of A. I can't think of a reason not to do it in the other direction as well: given that we have data set A, we should really question data set B as well. Unless of course exactly one of the data sets agrees with our preconceived notions, in which case we should just dismiss the data we don't like :P
It's more about seeking and analyzing the right data set to answer the question at hand and making sure that they are representative of the population you wish to study. If you'll allow me to elaborate:
Let's say we wan't to answer the question of whether there is a difference in the rate of homelessness when comparing ethnicity. The data being used to answer this question is a federal-wide census that aggregates the individual state census data collected over the year by each state. Let's say also that the statistician responsible has ignored the proportion of each ethnicity within the population of the state and has only reported an absolute number of homeless persons by ethnicity (e.g. State X Census: Caucasian 45k Homeless, Latin American 33k Homeless, African American 46k homeless, etc.). Once this data is aggregated, the same absolute results are used to represent the entire population of the US. This leads to any statistics derived from this data set to inaccurately represent the *proportion* of homeless people by ethnicity (i.e. adjusted for the size of their individual strata). Since only the absolute numbers are being observed, this can also lead you to make false claims about the populations you are studying.
This is called sampling bias or sampling error.
Indeed - she was smuggling her political bias into the talk.
Yeah, it's not like there could be some kind of "bias" affecting how police deal with a group of people who earned equal rights status a few decades ago. Inconceivable.
Wow so good! Taking stats in grad studies and this was fun to watch - and eye opening..
Don’t worry about the crowd not being responsive; one tends to be that way when in awe.
Surprised UA-cam hasn't censored this one out in the age of COVID🤭
The same apply in the age of global warming doomsday
Wow, she was doing so well and then bombed at the end. She actually took a survey of personal responses for something illegal and used it to try and prove a point, without providing any context. She also forgets most of these arrests and charges are usually for or combined with other offenses and can get pleaded down to a single charge and many other circumstances including who is interpreting and writing the paperwork; Simpsons paradox anyone...
ABsolutely... shes so "generic" in all of her talk... and shes tryng to discuss "data" with her audience..LMAO
Yeah, she herself drew a causation using statistics without actually looking at the data. A fallacy.
She oversimplified the facts behind the troubling statistic at the end, perhaps to cater to the people who believe the justice system is skewed against minorities. It may or may not be but there is more to it.
Wow. She explains how to spot estadistica fallacies in the news. And at the end she pull one of those on the attendees. And nobody even noticed it. She used the exposure of a stadistical fallacy to validate another one.... Low
She is amazing, such a professional in spite of the feedback she's receiving. Perhaps the quiz to see if anyone knew what Technology, Entertainment, Design meant rubbed them the wrong way🙄
Poverty leads to crime, groups that have higher rates of poverty will have higher rates crime. It is less about racist police and more about economic disparity. That is to say the issue is not that the stats are over-inflated per se, it is that they are proportionate to rates of poverty which leads people to criminal activity and rates of poverty are not equal among all groups.
That's exactly what she was talking about. That's a good hypothesis you have there, but do you have any data to back it up? It's easy to come to conclusions and even believe them to be true, but are they?
Everytime someone quotes a statistic I ask myself.
Where are the studies behind the studies?
I'm currently ending my 4th semester at political science (where we also learn statistics) and it's HAAAAAAAAAAAAAAAAAAAAAAAAAARD as @$%# to prove that anything has causality. Mostly the goal is just to reduce the chance of there not being causality by a lot and then assume that there's causality.
And this is what even PHDS do and our leading scientists. They're just so hardcore that they can reduce the chance of failure by very much but they still can't guarantee causality.
More meaningful discussions of causality are to be found in the field of philosophy - not political science (scarcely a science at all).
Stephen Powdexter political science is the corrupt mans philosophy. That’s an oversimplification of course, but it can’t be denied.
@@drewpocernich2540 If it can't be denied, then you can't causally infer the link between political science and what defines something as a corrupt man's philosophy :p
Dave Cullins your right, I guess I meant the political philosophy of boss tweed which I think we could agree on.
now apply this to covid stats, ridiculous from the start😂
Just to be clear at 5:06, there is a 12% increase in abortions and a 52% decrease in cancer screenings from 2006 to 2013 according to that set of data. Plotting things on the same scale is fairly irrelevant. I would hardly say 12% increase is "unchanged."
Her presentation was excellent! I don’t know why the audience was so dry, but I wish she would have the chance to do her presentation at another TED location. Outstanding job on her part.
Ok til the end when she felt it necessary to editorialize about a huge topic, law enforcement policy and behavior, had to drop it from my recommend list.
Glad to see others here in the comments section saw the same thing - over-exuberant social justice warrior - lost cred.
May I ask why ? Just curious
Here's a stat that isn't a lie: 100% of this audience were as dry as the Sahara
Anyone else waiting to hear the "results" of the poll she did in the beginning? Was expecting her to say something like "only 80% of TED Talk listeners have ever heard about TED" or something, since a lot of people for one reason or another didn't hold up their hand.
lol. Yes, I was expecting the results to be told, but I think she avoid it for not making the audience look stupid. She rather keep silence.
I didn't know the full form of TED so I wasn't sure, so probably the same with others too.
As Winston Churchill once said: 'I only believe in statistics that I falsified myself'
What are the five categories of lies she presented?
Several times in this boring and unnecesary talk, she points out somebody lying with statistics, and then posits an equally unsupported explanation rather than just leave it at there is no evidence for that.
Here's a fact to take seriously...
"The Huffington Post doesn't do honest stories."
**drops mic**
Her representation of the processed meat increasing bowel cancer risk by 20% is also a misrepresentation. One factor she doesn't consider: population number. So while your personal absolute risk goes for 4.4% risk of bowel cancer to 5.2% risk of bowel cancer if you eat 50g of processed meat per day, over a population of over 300 million that .8% points is absolutely massive. 300m x 0.8% = 2.4m people
Why would population number be relevant here? The population won't affect this statistic, its still the same increase for everyone?
@@HHumii if each individual's odds of developing the condition goes up by 0.8%, you would expect to see 0.8% more people develop the conditions with a large enough sample size. If you don't see why intuitively (skip this paragraph if you do), consider that we can determine the odds of something happening by taking the number of times it did happen divided by the number of times it could have happened. For example, if you were to roll a dice 600 times, about 100 of those rolls will be 1 - you can determine that you have a 1/6 chance of rolling a 1. The reverse is true as well - if you know you have a 1/6 chance of rolling a 1, you can expect roughly 100 of any 600 rolls of a dice to be 1. This applies here, too. If the odds of developing cancer in each individual increases by 0.8%, you would expect to see roughly 0.8% more of the population develop cancer.
In a large population, if the total number of people developing bowel cancer increases by 0.8%, though it sounds small, can be a very large number of individuals effected. 0.8% of the US population is around 2.4 million people - that's 2.4 million people who otherwise wouldn't have gotten bowel cancer. That 0.8% can be massive on the scale of an entire country.
@@HHumii It's relevant on the scale of the whole population, which is usually what these statistics are about. I think it's really weird how she dismisses a 20% increase (that actually IS a 20% increase!) because the numbers "only" go up from 4.4 to 5.2% - that's a HUGE effect on a national scale (let alone an international scale) and honestly, it's not irrelevant on a personal scale either. What she didn't mention is that, the incidence of other types of cancer and cardiovascular disease go up with the consumption of red/processed meat, so yeah, cutting out processed meat, can make a huge difference for an individual as well, over the course of their lifetime.
Uh that was uncomfortable watching it I could only imagine being there, she really should have showed the results of the 3 questions she ask the audience before hitting those high level touchy subjects
audience was mid
1.Statistics are more involved in making choices in our daily lives than we think. Therefore, we must look at the statistics carefully
2. Because statistical graphs are easy to manipulate, the source of these figures should be investigated.
3. Statistics can change over time because a lot of information is gathered and extracted
4. When interpreting statistical data, consider whether there were other events at the time of the event for that subject.
5.Misinterpretation of statistics can hurt others. Therefore, we should look at the statistics from a critical perspective.
Her voice is in the grey area between annoying and soothing
Fantastic presentation! Great performance! Thank you Sanne!
I was on holiday in the USA and was waiting for my wife outside a shopping mall.
A young lady approached me and asked if I would mind answering a few questions on the products for sale. My answer was 'No thanks' I'm a tourist here and I have no experience of using American products.
Her reply was 'That's quite ok, but you get 6 Dollars for completing the questionnaire and it takes just 4 minutes to answer. I then asked her if she was earning commission on every completed questionnaire to which she replied 'Yes of course'
Is it really any wonder that just hearing the word 'Statistics' is interpreted as an exercise in deceit and dishonesty to anyone who thinks for themself.
The last explanation of the drug offenses doesn't really make sense. Taking drugs is just a small portion of "drug related offenses" (selling, delivering, consuming). But since her explation is just about the consumption, it doesn't prove any racism.
Nice and early
Very good. Great way of explaining the numbers. Pity she went for the easy answer at the end of racism spoilt the whole thing
Judea Pearl’s “The Book of Why” would be an excellent basis for a short course in defending oneself against misinterpretations of statistics, and even better how to better interpret them.
Wow. She explains how to spot estadistica fallacies in the news. And at the end she pull one of those on the attendees. And nobody even noticed it. She used the exposure of a stadistical fallacy to validate another one.... Low
The US law should raise the standard of statistics put out by the media. And I don't say this lightly as generally a smaller government type.
13:00 MS Word is my bet
One biggest manipulation is omitted here. Magic of the Average (average of 99+1 is still 50)
Something is wrong and don't blame the audience! Sanne is trying to be natural, but it's all calculated and she's unnatural.
Something in the atmosphere sucks. It's a fail!
Sure she got some knowledge and it's an interesting subject, but being politically correct and nice in order to have her head patted is just lame (ref. chapter 2 of her book).
Thank you. Very good introductory talk help us better see and arm ourselves with critical thinking skills to help defend us against the damned lies held in some media driven statistical data.
Thought the ending was good too and needed saying. Brave woman.
By being reasonable!
Good job, good warning !
dream: interesting
im glad she brought up the integrity of the data as well, because even with these statistical differences there could be bad data to start with like including anything near school with a loud bang noise as mass shooting lol. many others of course but you get the idea media is really good of making things ten times more crazy
"Newer - Betta- DATA"... I love way she points out that not ALL things come from an opinionated reporters "hard data"... you can eventually prove any of your points , just find the right data..LOL
Best lecturer ever.
Thanks
How to prove a point about the fallacy of using statistics without context to then use a statistic without context to disprove another statistic also without context while also drawing conclusions.
So interesting! Thank you!!
Now we need a video titled: "How to defend yourself against misleading titles on UA-cam". When you say "How to..." there should be an actionable task to follow.
simple: always read the original research papers instead of news articles and be *extremely* careful with your conclusions
She told you how to: Do your own research into what the numbers actually mean.
She needs to come and do a statistical study on drugs and crime show her work and give a conclusion.
Mark Twain was skeptical of statistics. Smart man - and funny.
She forgot to mention that each one of us is a statistical anomaly.
poor tom cruise lmao, she has 0 mercy on short guys
reductio ad absurdum fallacy
There are lies, dam lies and Statistics.
poor people wasted there time
She abandoned data and any credibility at the end .
5 lies:
1. 3:51
2. 5:24
3. 7:26
4. 10:00
5. 12:10
She's making fun of short guys.
5 lies:
1. 3:51
2. 5:24
3. 7:26
4. 10:00
5. 12:10
Her #5 is flawed. Correlation ≠ Causality? If this were always true, then there'd be no such thing as science. It's more accurate to say that correlation does not guarantee causation.
"Correlation IS DIFFERENT from causality" does not mean that there are no causal correlations. Is like saying "swans =/= white", it does not deny even that there's a majority of white swans.
@@petitio_principii And you just illustrated a perfect false correlation between your swan example and the original argument.
Every effect in science is an instance where correlation = causation. This is very important to understand. But when you have virtually no correlation (except the accidental variety), you can have NO causation. Example: a rise in UFO sightings coincided with a sharp rise in the stock market. This is an accidental correlation, not a cause-and-effect correlation. But when you pour two chemicals together and get a pink liquid, every time you pour the same two chemicals together, that's a cause-and-effect correlation. It's how the universe works!
Hope this helps.
@@RodMartinJr correlation itself never "equals" causation; it can only be _"caused by a causal relationship."_ To say they're the same is simply a categorical mistake. Correlations are only "clues" to eventual causal relationships, you can have causal relations masked by other factors that will result in an absence of visible correlation or even a _negative_ correlation between cause and effect.
When you can filter those factors out, the correlation will be seen, but it is still true that correlation is one thing (co-incidence) and causality is another (the actual mechanics that should produce non-random patterns, correlating cause and effect).
@@petitio_principii Well said. I agree. You're using more precise language, and I love it.
But I stand on my first statement. Too many people disconnect correlation and causation, missing the most important connections that are the foundation of science.
Take the "Climate Change" scam, for instance. Sure, there can be other things masking the CO2 effects on warming. I don't doubt that for a minute. But CO2 is pretty wimpy for its "signal" to be invisible on virtually every time scale but one -- and in that one, temperature is driving CO2 into and out of the oceans.
The warming of the last 300 years and the increases in CO2 are largely, if not only, an accidental coincidence. The Modern Warm Period is the *_Coldest_* of the Holocene's 10 major warm periods (1,000-year cycle), and CO2 couldn't keep the Modern out of *_Last Place!_*
That is absolutely true, evenmore so in some cases a correlation of zero guarentees independence etc. However the problem is that in the real world the statistics requires one to make assumptions and infer a distribution which can be incredibly hard. Most journalists and whatnot have no clue how immensely complex statistics actually is
first
3:19 What are they doing there?!?
Who the heck are they?
great talk very honest cuts straight good speaker
Excellent!
LEIP
la caquita
Kurrino
Lovely; u r gifted.
wow!! keep up the good work Sanne Blauw!!
Feel sorry for her. She didn't have "it". Maybe this was someone else's presentation.
Great speaker, bad audience. Who'd she have to follow? Zombie Steve Jobs?
I noticed something-again. At 14:40 she presents drug arrest numbers for whites and blacks: 332 vs. 879 (out of every 100,000). At 15:23 she presents illicit drug use statistics: 9.5% for whites, 10.5% for blacks, a difference of 1%.
Blacks use more drugs than whites according to the data she presents. Then we should expect the arrest rate to be higher for blacks than whites.
Now, 1% of 100,000 is one thousand. If every drug user was arrested, should we then expect numbers of 9,500 for whites vs. 10,500 for blacks? I don't think so-if the arrest figures are yearly, if every illicit drug user is immediately arrested and given a two year prison sentence (and they don't escape), we should see those large arrest figure in one year and 0 in the next (for both racial categories), or maybe you'll have one half of druggies in prison and the other half roaming free, with the two groups switching roles every year, so you have stable arrest figures of 4,750 for whites and 5,250 for blacks. And in reality the arrest rate is less than 100%, and so on and so on, there are many complicating factors.
But if the drug use rate is higher for blacks, I would expect the arrest rate to be higher for blacks.
Maybe the arrest rates should be proportional to the illicit drug use rate; 10.5% is higher than 9.5% by about 1%, which is approximately 10% of 9.5%. So the arrest rate for blacks should *maybe* be 332+33.2 = 365.2 (per 100,000), and we can argue about the reason for the gap between 365.2 and 879. Maybe 365.2 is wrong for subtle reasons. But this stuff should be argued in considerable detail before you jump the gun and conclude "MUH RAAAAYSISUM".
She claims the drug use is the same for the two ethnic groups. I assume this is because the study giving us the 9.5% and 10.% numbers show a difference which is not statistically significant. In which case my analysis is somewhat moot. But then she should explain why she displays two different numbers yet says there's no difference.
Also, how to lie with statistics #637: present two sets of data using different units, such that the absolute numbers are very large in one case (332 and 879) and very small in another (9.5 and 10.5).
I was waiting till the end for the wage gap myth busting. She did well tho - one truth at a time.
Yet she jumped to an unsubstantiated conclusion by calling police racists. Perhaps there is another answer, but she didn't want to pull the data to check. She turned a decent talk into a thumbs down with her unwarranted political commentary at the end.
Thought it was Natalie Portman in the thumbnail
This is quite relevant now.
Good video but a nervous presentation.
Why do we need an "ality" on the end of cause?
This audience seems to hate this woman.