Yuma 4×4

Media and Communications

How to spot a misleading graph – Lea Gaslowitz

How to spot a misleading graph – Lea Gaslowitz

A toothpaste brand claims
their product will destroy more plaque than any product ever made. A politician tells you their plan
will create the most jobs. We’re so used to hearing these
kinds of exaggerations in advertising and politics that we might not even bat an eye. But what about when the claim
is accompanied by a graph? Afterall, a graph isn’t an opinion. It represents cold, hard numbers,
and who can argue with those? Yet, as it turns out, there are plenty
of ways graphs can mislead and outright manipulate. Here are some things to look out for. In this 1992 ad, Chevy claimed to make
the most reliable trucks in America using this graph. Not only does it show that 98% of all
Chevy trucks sold in the last ten years are still on the road, but it looks like they’re twice
as dependable as Toyota trucks. That is, until you take a closer look
at the numbers on the left and see that the figure for Toyota
is about 96.5%. The scale only goes between 95 and 100%. If it went from 0 to 100,
it would look like this. This is one of the most common
ways graphs misrepresent data, by distorting the scale. Zooming in on a small portion
of the y-axis exaggerates a barely detectable difference
between the things being compared. And it’s especially misleading
with bar graphs since we assume the difference
in the size of the bars is proportional to the values. But the scale can also be distorted
along the x-axis, usually in line graphs
showing something changing over time. This chart showing the rise
in American unemployment from 2008 to 2010 manipulates the x-axis in two ways. First of all, the scale is inconsistent, compressing the 15-month span
after March 2009 to look shorter than
the preceding six months. Using more consistent data points
gives a different picture with job losses tapering off
by the end of 2009. And if you wonder why
they were increasing in the first place, the timeline starts immediately after
the U.S.’s biggest financial collapse since the Great Depression. These techniques are known as
cherry picking. A time range can be carefully chosen
to exclude the impact of a major event right outside it. And picking specific data points
can hide important changes in between. Even when there’s nothing wrong
with the graph itself, leaving out relevant data can give
a misleading impression. This chart of how many people watch
the Super Bowl each year makes it look like the event’s
popularity is exploding. But it’s not accounting
for population growth. The ratings have actually held steady because while the number
of football fans has increased, their share of overall viewership has not. Finally, a graph can’t tell you much if you don’t know the full significance
of what’s being presented. Both of the following graphs
use the same ocean temperature data from the National Centers
for Environmental Information. So why do they seem to give
opposite impressions? The first graph plots the average
annual ocean temperature from 1880 to 2016, making the change look insignificant. But in fact, a rise of even
half a degree Celsius can cause massive ecological disruption. This is why the second graph, which show the average temperature
variation each year, is far more significant. When they’re used well, graphs can
help us intuitively grasp complex data. But as visual software has enabled
more usage of graphs throughout all media, it’s also made them easier to use
in a careless or dishonest way. So the next time you see a graph,
don’t be swayed by the lines and curves. Look at the labels, the numbers, the scale, and the context, and ask what story the picture
is trying to tell.

100 thoughts on “How to spot a misleading graph – Lea Gaslowitz

  1. I thought it was interesting how at the end, she defended the misleading graph. The honest graph shows a fairly consistent temperature over the span of 200 years. The alarmist graph takes a much smaller span of time and exaggerates what is really just statistical noise.

  2. But the Chevy truck ARE more dependable than the Toyota trucks! if you divide 100 by 1.5 (98.5% of Chevy trucks still on the road) we find that one in 66.6 trucks will fail, whereas if we divide 100 by 3.5 we find that one in 28.6 Toyota trucks will fail, meaning that the Chevy trucks are just more than twice as reliable!

  3. 1:23 I do agree. But a 97% approval rating still means you have a 1/30 chance of getting screwed over, while a 95% one means a 1/20, 96,98,99% = 1/25,1/50,1/100. So yeah, it's more like the look at it the negative implication & flip what gets filled.

  4. Also, don't assume a correlation just because two graphs seem to match. If the match isn't great, like the famous "lead versus crime" graph, there's a good chance that it's just represented in such a way as to highlight a coincidence. Even if the maps follow in eerie lockstep , it's still possible for them to have a non-causal relationship, like the way that ice cream sales correlate with violent crime but both are just promoted by heat.

  5. Well, like my maths teacher told us: "never trust a statistic that you haven't manipulated yourself"

  6. For the car ad one, your car is half as likely to be off the road in a bunch of time. The "Barely detectable difference" is what MATTERS.

  7. Oh this is perfect.Alright you hypocrites, here's an important piece of information about the earth's temperature that you've left out:The chance in carbon level happens up to a year before the change in temperature. That's why they're always zoomed out so much. Also, in 1880, we didn't have measurements accurate to within a tenth of a degree, so to say that average temperature has risen by that much since then is preposterous. The idea that slightly less than 13ppm of carbon is causing massive changes to the planet is both laughable and thermodynamically impossible.

  8. I mean the truck ad is stil technically correct the chivies are half as likely to break down as a toyota. It comes down to how relevant the doubling is, if the ad was how many cars hadn't had accidents then the difference between 98% and 96% success rate is pretty massive.

  9. Back in elementary school we went over this, and one assignment was we were given a bunch of data for some companies (I think it was sales made), and we had to make a graph to a demand for each company (Company A wanted it to look like a landslide. Company B wanted to look like they were close. etc.)

  10. It's good things to be aware of even if not to avoid being manipulated, then to not manipulate others by accident

  11. You know, you made some good points, but it's hard for me to respect you when you condemn "cherry-picking," then turn around and engage in cherry-picking yourself. Namely, in which side you choose to target when selecting which "misleading graphs" to use in your examples.

    No prizes for guessing which corner you're batting for.

  12. 1:41 Hmm imagine that, Fox News using a misleading line graph. Who would think they would do such a thing.

    They probably have equally as bad graphs comparing Trump’s job growth to Obama’s. Obama’s probably looks like a pancake and Trump’s like Mount Everest.

  13. At 2:07 , they showed the unemployement rate. Okay it hasn't increased, but has it declined? Unfortunately no. The left has not taken people out of poverty. Until Trump came, unemployement of blacks has been at it's lowest as well for women, hispanics, latinos.. etc. So Ted-ed has made no point

  14. Car manufacturers are the worst with manipulating stats to their favor. Especially with trucks. "Biggest payload capacity in it's class!". That's great, but then how does it rank in the other 5,000 features you can rank it on?

  15. And Bhartiya Janta Party of India goes to another level:

  16. This is like analysing sources in history class and practicing making them in math and science, high school doesn’t seem so useless now does it

  17. 2:59 "Annual Glob Oc. Temp". is actually more consistent that the "Ann.Glo.Oc. Temp.Anomalies"on a highly reduced scale of ONLY -1 to +1 degree as to hyper-accentuate otherwise insignificant variation. "ANOMALIES" is a highly PARTIAL and CHARGED term, because the second graph has an AGENDA to mislead the audience – just as it was done by GM in the chevy truck graph 1:02

    3:15 the narrator says "avg. temp. each year," when the graph clearly shows a different TITLE: "avg. glo. oc. temp. ANOMALIES." The video itself is CHERRY PICKING and disingenuous.

    ANOMALY, means "not normal," but in continuous evolving biosphere, the term norm.temperature, norm.average, or norm.climate is MEANINGLESS and MISLEADING. We can have an "average ~ norm" in a short time range, like from 1880 to 2016, but if you increase the scale, the temp.curve will be all over the place. ex.: from 10,000 ya to today. For 3.5 billion years of earth, "CHERRY PICKING" of 1880 (beginning of weather records) is statistically insignificant.

  18. The news ticker tapes:

    "Cat Floating in Ocean Rescued By Coast Guard"
    "Just Lukin' Renewed For Three More Seasons"
    "President Thumb To Congress: 'Think of it as a New Rule of Thumb'"


  19. 2:34 superbowl **releases super bowl halftime show 2019**v -3829478268462374723423847249734895734897638563478568346974628463578246592456947825623489576458237659456498725649546295624579167050571467504156107876507461047817813063580641084561078888836078138060481561408561485648888888888888888888875714514056445151513076107804136504357105834165078104095819347051780528579847508374985720453720f89574893572057489-1-8104709137054387061574308160534605436054307810534785 views

  20. They managed to miss the point on the ocean temperature graph. START at 0. For a second there, I thought they might tell the truth, then they supported the lie.

    If you start at 0, you will always have a clean picture. If you make up a 0 (like they did in the chart magnifying the 1/2 degree change), you will be dangerously close to fooling someone.

    The first person to be fooled is YOURSELF.

    Always start with a chart that starts with 0. Know what the zero means and how to arrive at its definition. 0K is a really good 0 to use. We know how to get to that value. 0C is also a good zero to use, we can replicate in our homes with just a cube of ice and a glass of water.

    0 C Anomaly is a terrible number to use. Calculating anomalous zero is fraught with ways of fooling yourself.

    Correcting past data is fraught with ways of fooling yourself. Especially if you correct it algorithmically. Those algorithms can make it so a temperature today will change the temperature 120 years ago. This should scare the living daylights out of people. How can a temperature in the past change because of the temperature today? Peak at the algorithms and you find out that they bracket the temperatures with years ahead and years back. Every temperature is bracketed. It should not happen, but some clever person made it happen.

    Start at 0. Stay linear. There is a place for logorithmic charts, but it is with day to day analysis.

  21. In France, a major TV news channel showed a pie graph with 47% of people not supporting a march against a new labour law… with 2/3 of the graph colored. The title was: "The majority of French don't approve this march".
    and they showed 32% approved the march with roughly 1/3 (which is honnest) and the 21% "without opinion" was a tiny slice.

  22. So we're all gonna ignore the fact that the guy saw the graph at the bar, examined it so intently that his eyes fell out and then married it? Is this some kind of subliminal relationship advice?

  23. A good book to read is How To Lie With Statistics by Darrell Huff. An old book but still relevant. There is another book used in an Uol Stats undergrad course but I can not for the life of me remember its name. It is lost in my room under some clothes or books.

  24. @3:14 this graph is more important because it uses a much smaller time segment, you know don't be miss led by stacked statistics

  25. Of course a Fox News graph made the cut… Fox manipulates data all the time as part of their relentless propaganda.

  26. I think the first example of the chevy exaggerating their graph was not appropriate, because it was about the rate of durablity of their trucks. Surely, they exaggerated the graph, arguing theirs are the best. But it doesn't mean that the mere percents of difference doesn't matter. Assuming it was the rate of trucks broken, we can say toyota trucks almost doubled chevy in loss.

    I think it's almost like a fielding rate of fielders in baseball. Between 96% and 98% seems small difference, but it says the 96% one commited twice more errors than the other.

Leave comment

Your email address will not be published. Required fields are marked with *.