How to defend yourself against misleading statistics in the news | Sanne Blauw | TEDxMaastricht

that's it's so great that you're here I'm doing this study and I just need a couple of more subjects so can I ask you a couple of questions yeah great thank you thank you so much okay just raise your hand if it's a yes and keep your hand down if it's a no super easy yeah first question have you heard of Technology Entertainment Design t-- e-- d anyone i see 90 keep your hands up please 93 points 9 3 4 % yeah great um second question do you have the flu but like really bad high fever can't get out of bed flu 0% wow I guess Tet must make you healthy right very interesting this impressive um last question this week have you seen a number this can be anything I'm a percentage an average a graph have you seen a number in the news quite a lot thank you this was great really helpful so let me introduce myself my name is song a blow and I am a number nerd and my addiction started when I was quite young in fact one of my very first memories was about numbers remember these connect the dots these actually did myself when I was about four years old it was on this vacation with my family in Germany I was just doing one after the other to the other super addictive then fast-forward 20 years I was doing my PhD I was collecting data analyzing data and still in a very different way but still looking for these patterns and numbers and now I'm a journalist and I write about the value of data and statistics and I think it's just that that we can see certain patterns and those patterns can teach us something about the world so it's no wonder that our news is full of numbers whether it's some headline about refugees coming from an international organization whether it's the outcome of a political poll or whether it's some groundbreaking new health insight based on statistical analysis smelling farts might prevent cancer before you get too excited this one turned out to be bogus so behave yourselves please also up there great um but whatever they are these numbers they shape the way we view the world they tell us what to eat they tell us what kind of medicines we should take when we get ill they tell us how the economy is doing and who we should vote for so it's very important that these numbers that they are correct that we can trust them sadly though many of those statistics in the news that are supposed to inform us they actually miss inform us as a wise man once said torture numbers and they'll confess to anything in our news today sadly is full of such tortured numbers so I think it's time for some statistical self-defense today I want to talk about the five most common statistical lies that you see in the newspaper every single day right ready great one the good-looking graph you know these graphs that you see in the newspaper on the news they just look too good to be true and quite often they actually are a bit too good to be true take this one this Gareth was shown in the American Congress in the midst of a huge discussion about Planned Parenthood this Health Organization and one Republican congressman put this graph up and he said look Planned Parenthood is doing less and less cancer screening and prevention services the blue line is going down and the purple line abortions are going up dramatically quite steep so we should we should cut funding to this organization because why should we spend our taxpayers money on things that aren't saving lives something weird with this graph though because look at the numbers the upper right hand corner says three hundred twenty seven thousand below it says nine hundred fifty thousand and a little bit don't really need a PhD in statistics to know that lower number should be lower on the graph so what happened here this graph uses two different scales if you put them on the same scale the message is very different okay cancer screening and prevention is still going down but abortions have hardly moved seconds disco line the polluted pole but we already did quite a polluted boulders now I think but there are many other polls out there in the news this one for instance one in four women experienced sex assault on campus this was a headline in New York Times and it's incredibly shocking right twenty five percent of female college students have to deal with sex assault fortunately though it's also probably exaggerated and the researchers admit this they say there's probably a bias upwards but the New York Times didn't mention this in the article so what is the issue here well three things one a very small sample of universities participated to at those universities more than eighty percent of the women did not respond to the questionnaire and three the definition of sexual assault was a very broad one here in fact the researchers didn't even were they avoided the word assaults because it's a very loaded term so what could happen is that a girl at a party being hit on by a guy and rejecting him she might be Coins evict by this survey even though she would not cool that Assault herself so someone in The Huffington Post said hey maybe the New York Times should have had this as a headline approximately one in four of 90% of non represent a sample of women who've responded to an armed center servile turns after God reported experienced sexual assault with sexual assault take to make anything from being on the receiving end of Iran so this gives first of all penetration gunpoint regardless of particular context I hope you're still awake because this is not really clickbait material and but it is the more on a story here right and I think when we talk about something as serious as sexual assault we should also take the facts seriously so third statistical I'm the overconfidence decimal points our news is full very precise numbers with two or three digits behind the decimal point and that gives the impression that these numbers must be true objective because they're so precise well let's take a look at one the kind of the queen of these things GDP growth GDP growth is an important number because based on this statistic economists tell us whether our country is in recession or not so last year statisticians in the United States announced that April to June 2015 the GDP growth was two point three percent a month later said oh sorry three point seven percent another month later said oh eight three point nine percent so it jumped so much in such a short period does this mean that these statisticians were not doing your job well no what this means is that GDP is incredibly difficult to calculate and it takes time just look at this very boring table don't even bother reading it but just to show you only components that go into GDP they're a lot so it takes time to get good accurate data on these no wonder that these statisticians were revising the numbers as they went so you might wonder like why should I care I never liked economics in high school and well this probably didn't make you more excited about it so why should you care now right well let me give you an example two researchers looked at 40 years of data in the United Kingdom and they took an older data set and they concluded there were 10 recessions in those 40 years with a better data set so with newer better data they concluded there were only seven so three whole recess recessions disappeared simply because they got better numbers and these were periods in which people were fired in which politicians won elections over over bad economic circumstances periods in which countries were in complete chaos because of this supposed recession so maybe you should care okay fourth the spectacular statistic or I should say they're not so spectacular statistic this is kind of the Tom Cruise of Statistics you know in movies he looks impressive right Wow well the suit really helps as well but still you know Tom Cruise he always looks awesome and powerful and everything put him in context so yeah yeah I prefer the previous one anyway the same goes with statistics sometimes you add a little context and they look very different take this one for instance people who eat processed meat daily have a twenty times higher risk of getting bowel cancer this was the headline in the Dutch news well a little bit after they said oops sorry it was 20% higher quite a difference but but still 20% still a high number right at least high enough for a lot of media here and in the United States and in other countries for a lot of media to write about this but the right question here is 20% of what because 20% of a small number is still a very small number so let's take a look in the United States now the chance to get bowel cancer is about four point four percent if you eat some extra meats 50 grams which is about one hot dog feed a hotdog per day extra that number goes from four point four to five point two percent already quite less impressive than that big 20 percent we saw before even less impressive it gets when you look at the chance of not getting cancer because that is ninety five point six percent in the first case ninety four point eight percent if you eat this extra hotdog per day so important information if you decide about your diet but not as spectacular Tom Cruise II as we saw before last one the cocky correlation the fact that two things happen at the same time or happen through the same people that doesn't mean that one causes the other let's take a look at this graph annual incidence of brain tumors this is in the United States and it goes from mid 70s to mid 90s you see something weird happening in mid 80s you see all of a sudden this graph shifts upwards it's quite a big shift and the researchers that publish this graph said hey what's going on here and they have an answer they said it must have been this NutraSweet this artificial sweetener that was introduced in the United States just a couple of years before this big shift in the graph we know what else happened in the mid 80s I don't know band-aids Microsoft Word I have been mid 80s so a very life-changing events obviously but but still without any additional information I am just as likely to have caused a big increase in brain tumors as NutraSweet you know what there was a much more plausible explanation and that's this MRI scanners before some raid tumors just went unnoticed because they didn't have the instruments to actually detect them and now mid 80s more and more MRI scanners were around in the United States so more tumors were actually noticed so this is only one example where correlation is being mixed up with causality and to me this is kind of the most dangerous statistical lie of the five because it's so easily linked to action once for instance I tell you that TED talks are super healthy you might decide to just you know spend all your time sitting in theaters like this and watching TED talks all the time because you want to be super healthy you might even quit your job but this a silly example of course but look at this one this is a graph from voxcom and it's about drug-related arrests in the United States and you see that black Americans were more than two and a half times as likely to be arrested for drug-related offenses than white Americans and the easy conclusion ears is hey well being black must make you a criminal right because look look at the stats and then the action could be obvious maybe the police should go to black nerd neighborhoods should single out black individuals because that's where the bad stuff happens right we see it here in the numbers but look at this one black Americans are not more likely to use drugs they're also not more likely to sell drugs so that correlation that should be second-guessed I probably the whole reason that we see that big difference is those racist police practices in the first place so I think we should defend ourselves against this cocky correlation and also against all the other for statistical lies I talked to you about because if we don't watch out these bad statistics they end up hurting ourselves hurting other people and hurting society so thank

Channel: TEDx Talks

Views: 123,230

Rating: 4.7606468 out of 5

Keywords: TEDxTalks, English, Netherlands, Social Science, Education, Journalism, Research

Id: mJ63-bQc9Xg

Length: 16min 20sec (980 seconds)

Published: Thu Nov 03 2016