When Numbers Lie
Вставка
- Опубліковано 5 сер 2014
- Facebook: / frameofessence
Twitter: / frameofessence
UA-cam: / frameofessence
Sources:
Causality: Models, Reasoning, and Inference
books.google.ca/books/about/Ca...
Internet Project
wps.aw.com/wps/media/objects/1...
Visualizing Urban Data ideaLab
vudlab.com/simpsons/
Can Quantum correlations be Explained Casually
pirsa.org/13070030
Images:
Casimir Effect
commons.wikimedia.org/wiki/Fil...
Cosmic Microwave Background
commons.wikimedia.org/wiki/Fil...
Atmospheric Carbon Dioxide
commons.wikimedia.org/wiki/Fil...
Music in this video (from the UA-cam Audio Library):
Fortaleza
Good Starts
Zydeco Piano Party
So numbers don't lie. But we can misunderstand them
Maybe the title should have been when numbers seem to lie.
I’m over here, like, clearly the problem is the placebo. Take that irrelevant variable out!
maybe the fucking title should be "when I lie for clickbait because I'm that much a fuckwit"
"There are three kinds of lies: lies, damned lies, and statistics." - Benjamin Disraeli
+Picotry Nice, I like that quote :)
Just found your channel today (December 12th, 2017), and I'm binge-watching. You're videos are exactly the kind I like. Intellectual and informative about topics I enjoy.
Thanks for your videos! You're one of my new favorite UA-camrs.
I really enjoy your videos. Please keep making them, I'm sure your fan base will grow a lot in the next little while.
Thank you! :)
So, don't hate me, but doesn't simpson's paradox apply to the wage gap?
+LarlemMagic absolutely
+LarlemMagic *ding* *ding* *ding* We have our winner :D
+LarlemMagic of course, it's just that debunking what's probably the biggest hoax in today's western world in order to show an example could be quite distracting :-)
+Scapegoat how does that apply exactly?
+fahadAKAme Departments with higher acceptance rates are like higher overall paying jobs.
I'm not saying anything about weather or not the wage gap exists, but if you were studying it you would ABSOLUTELY need to account for Simpson's paradox. It would destroy your data.
If all women were payed $1 for every dollar men made in each field, but more men were in higher paying fields of work, then your overall data could show men making more.
Don't hate me though. I'm not saying the wage gap doesn't exist, just that any study investigating it would need to account for the paradox.
...I have three rocks
Do u
3:11 "(hopefully...)" - That is not good enough for me.
4:26 Great example. And I think that this type of paradox is far more common than what we've been led to believe (accidentally or otherwise).
paradox? you mean mistake/human error?
I'm so glad I discovered your channel! Thank you for making these videos, the animations explain well and I've learnt new and interesting things. :)
Thank you! I really appreciate it.
Hm, perhaps important to note is that the odds of getting 150 of the same gender on one half of the randomly selected group, is less than 0.0000000000000000000001
If you select from a group purely randomly, you wont be exposed to Simpson's Paradox with any reasonable level of possibility. And the existence of the possibility of Simpson's Paradox has no effect on the data, other than perhaps giving a bimodal distribution which would give you a poor SD if you assume it's unimodal (Which, a poor SD is not a bad thing, it just means you have the opportunity to make your deductions stronger by splitting by gender and then analyzing).
The case of Berkeley was not random, of course.
You deserve more subscribers! Just saw your channel because of the response video you made to veratasium, cool stuff
Thanks for the explanation. I have read the wikipedia page multiple times but never really understood the simpsons paradox. This video cleared it all up!
The problem with this is that how do we know what subsets (male/female in the drug example and departments in the school example) to look at? We can't possibly split every experiment's people into, "people who prefer brie over camembert" vs. "people who prefer camembert over brie," or "people who have at least three vowels in their first name" vs. "people who have fewer than three vowels in their first name" just in case cheese preference or vowel count might have an effect on something. It seems like, in general, there's going to be a lot of guesswork to decide how to break apart the sample into groups - and we might have a bunch of bad results from experiments that do so incorrectly.
+Viola Buddy Indeed that is the case. Makes medical studies very difficult to interpret and extremely prone to error.
Viola Buddy
If you know that women can more easily recover from a disease, then that means you know that the men and women should be given proportional amounts with men getting half and women getting half.
It's also worth considering that the probability of only 50 men getting the drug and 150 women getting the drug is extremely low. It's less than one in a trillion.
Getting exactly 100 men and 100 women is going to be about 0.05635 (5.6%); however, it's more important to consider ranges here.
For there to be 50 men or less (0 to 50) with the drug, the probability is 4.197E-13. For there to be either less than 50 men or more than 150 men with the drug, the probability is double at 8.397E-13. Remember that a trillion is 1E-12 and this number is lower.
The probability of 90 to 110 men getting the drug is 0.84182 (84.1%).
So you only have 0.15818 (15.8%) probability of even getting outside that range. You have less than a percent chance of getting outside the 82 to 118 range.
+Viola Buddy Sex difference may change the way body's chemistry works. There's little reason to believe that people's names change their body chemistry.
so with all that said, numbers don't lie just as much as cameras don't lie
Awesome channel!
THIS JUST BLEW MY MIND
I have to prepare a presenatation about data science and just got your work. Thanks you saved my life, I may be thief but thank you anyway
brilliant channel
I´ve seen this happening when comparing sales goals' fulfillment rate,
Great video
Nice channel!
Wait, why were the charges of discrimination dropped when they realized they were discriminating against men?
+DanielMasey I'd call a 4 percent gap too small to worry about. In other statistical situations it might even be smaller than the error factor. Also, when it comes to social stuff you should probably allow a fair amount of wiggle room.
As a man I do sometimes feel that discrimination against men is ignored, but I'm not gonna quibble over anything less than a 10% gap. Heck, I'll even listen to arguments justifying some discrimination, though I haven't heard any that hold up. On the same topic, there may be some apparent discrimination against men that turns out to be correlation not causation.
Timothy King If the revised data showed 4% higher acceptance for males the suit would have continued.
tbh, a similar trend is showed across most levels of education: Women and men are basically equal but only when fully grown. While they are still developing (which can take until the late 20s to finalize brain development), women hold a small but nontrivial advantage. This, coupled with societal perceptions against women taking certain trades assumed to be the domain of men that happen to not require college degrees (plumbing, construction, etc.), and suddenly you arrive at a weird situation where college educated women grossly outnumber men. Such overrepresentation is fairly recent, as many colleges previously either discriminated against women subtly or just tried to aim for 50/50 but now admit applicants gender-blind.
Roman Nasuti Gender blind plus special scholarships for women and oh wait
Because they are not women.. Man are not whiny bitches...
I wonder if Simpson's Paradox is the reason behind the voting results of the 2016 Eurovision Song Contest. Australia was leading in the Jury vote with the most points and Russia led in the Televote (public vote). But when the points from the Jury and Televote were added together, Ukraine won overall.
Place Televoting Points Jury Points
1 Russia 361 Australia 320
2 Ukraine 323 Ukraine 211
3 Poland 222 France 148
4 Australia 191 Malta 137
5 Bulgaria 180 Russia 130
How come you didn't just mention you can use a multi linear regression to control for gender using a binary variable? Like, I understand you you didn't have to go into OLS to make your point (well done, most of the audience would have run away screaming and unsubscribing) but any researcher worth salt with a first year stats course would just stick in a control variable.
Or, realistically, probably a whole tone of control variables.
I actually hadn't taken a stats course yet when I made this video, and I didn't want to risk getting any details wrong. I might include more details if I ever decide to remake it.
Wow. I just assumed you were a physics teacher, either at school or at a college.
Soooooo, not to be too subtle about this but I'd be happy to collab on such a thing, even though I realise that there basically no union between the our sets of audiences.
So the revision that corrects the data is to value each subset equally?
+Kieron George Pretty much. The tricky part is finding all the variables with possible causal relationships and correcting for those as well.
+Frame of Essence God, i love this channel, you da real MVP.
me clicking this video: hell yeah spill the tea on number drama
Mind blowing
Pretty interesting
Nice channel m8.
Cool channel!
interesting, thanks
That tau at the beginning though...
Tau for the win!
Why did u stop making videos
A RAMBLER MOLDY INSIDE -- CAN'T NEVER COULD DRIVE.
that' fucked up. Defo companies out there doing that 👀
CROSS SECTION OF GRID, AT THE CROSS, AT THE CROSS , WHERE WE FIRST SAW THE LIGHT. SCANNER
Soooo, stats aren’t a easy as we thought. Wait, we didn’t think they were...
What I learned: Numbers don't lie, stats do. (sometimes)
wait wait wait, why is it called a paradox when none of the statements conflict?
ua-cam.com/video/kJzSzGbfc0k/v-deo.html
This would never be an issue if the sample sizes could be arbitrarily increased to ever higher numbers.
IF IT HAD NOT BEEN FOR HER , I WOULD NOT HAVE HIRED YOU.
numbers lie it says 90% downloaded but it takes forever to download.....
I still don't understand it, maybe i should rewatch the video when the clock isn't 01:09 am
okay so why making a fucking error is somehow a paradox is probably because the paradox is specifically for combining data, which is troublesome because data naturally has its more complicated but vital information hidden away in the final answer, so if you try messing with data without fully understanding it youre bound to make mistakes for overlooking some maths
There is no paradox here you just added the fractions incorrectly. For example if you wish to add 10/50 and 90/150 you don't do (10 + 90) / (50 + 150) = 100/200 = 50% you first have to make the denominator equal 10/50 + 90/150 = 30/150 + 90/150 = (30 + 90)/(150 + 150) = 120/300 = 40%. Simple fractions can slip you up but the fault in your maths is odvious as the average of 20% and 60% is not 50% and like wise for 30% and 70% not being 40%.
Andy Adshead Yes, but if we had forgotten to account for gender in our calculations, we never would have that discovered that they were incorrect. We started out with the wrong totals, and we needed to find out that the totals were wrong by splitting them up for each gender. Who knows what other hidden variables we didn't find? If we missed any, then we may have added *them* up wrongly too. But how do we know we found every one?
Whoever set up the problem had to add the fractions together to get the numbers you see on screen, he added them wrong it's as simple as that. If I just added up the first 2 fractions and said they give 100% then everybody would just call me stupid so I fail to see why this carefully set up stupid maths is any different. Long story short if you tried to advertise with these numbers then your ad (at least in the uk) would just be pulled for lying.
Andy Adshead But the totals weren't found by adding up fractions in this example. They were found by simply counting the number of people with the real drug who recovered, and counting the number of people with the placebo who recovered. And you're right, this is the wrong way to do it, and there likely would be consequences if these numbers were published. As you said, you should look for all the hidden variables and add them with equal denominators for each one. But not everyone will notice a problem with the flawed calculations, which is what I find so interesting.
WHAT IS WITH THE NO TRESPASSING SIGN? DADDY PUT THOSE UP BECAUSE OF YOU.
UP AND DOWN JACOB'S LADDER.
UC Berkly does not give merit as a priority ? It prioritizes genders and sex ?
And so, when you look at the actual data instead of the totals, the gender pay gap... Disappears!
It decreases, it doesn't disappear.
Mark Neu If you’re looking at the U.S., then yeah, it disappears.
MOREOVERLY WE MANIFEST WHAT WE THINK EVER SO SLOWLY. THE GRID. CRYSTALLIZATION, SHOPPING BUGGIES, OPERATION THINK, IN THE TWINKLING, WHY WOULD ONE EMPTY? ONE , TWO, THREE, FOR THERE IS NO NEED.
REASSEMBLED MATTER
0:09 Can we take a moment to appreciate how he used Tau instead of Pi in one of the equations?
مصدر من مصادر الدحيح 😂😂
2020 here
0:11 Eulers identity is = to -1 not 1...
The equation has a tau instead of a pi, so it still holds.
Frame of Essence oh... thanks :)
no problem :)
i am sorry i sleep on my desk
e^(pi*i) doesn't = 1 wtf? lol
It's a tau, not pi.
true... thanks
JOHNNY WHITE, ARIZONA PHOENIX, CAIN' T NEVER COULD DR. CAN'T NEVER COULD DRIVE.
Says numbers Lie then says they don't
So minus the placebo, the numbers were true! 10/50% + 90/150% = 100/200 = 50%
مين اللي جي من عند الدحيح
انا
منظمة بص بصة عالمصادر
أنا يا كبير 😂
مطلعتش الوحيد اللى بيبص على المصادر
منورين والله
Sad little number wizard.
Well, the solution is to do AB testing.
what did we learn ?
That women are the stronger ones !
Wage gap post 1990.
so basically fixed statistics lie on numbers.
بص بصة عالمصادر
The drug don't harm gay
data indicate?! indicate?!?!??! triggered.
common sense is better than statistics
JAGLOIS JANINE
The charges were dropped? Why? They were discriminating against men.
Gaming
0:10 e^iп + 1 = 0
+Maksim Bandel It's a tau, not a pi.
that's not pi, that's the cyrrilic pe.
ESAU EVER SO STEADILY CREEPS.
اخوكم من عند دحيح
علقتلك ٣ مرات 😂😂😂
0:09 e^(i*pi)=-1 not 1
It actually says e^(i*tau) where tau=2*pi
+Frame of Essence oh yeah then its right :)
nice video btw
+Pieter Luyten Thank you :)
THE WARP, THE WEFT, THE WEAVE WE CREATE AS WE SPEAK . THE SOWER IS THE SEED.
Ohh like with the gender pay gap?
Grow up
"Statistics can be misleading" show some decency and change the title please children are watching!!
Luis Valentim But the final message clarifies it anyway. And the current title is so catchy.
You are right. I'm just so tired of misleading titles and lies everywhere... Please let me have numbers as the last honest thing.
Great work by the way.
Luis Valentim Thanks :)
+Luis Valentim Wait, so how is the title wrong?
Numbers don't lie
so numbers don't lie... clickbait title.
Steve Wil not really. What he said is actually a big problem in statistics.
First
Someone tell feminists this piece of info, so that they can realise the wage gap isn't real
They wouldn't listen anyway.
انا مش فاهم حاجة 😂الدحيح احلى بيتكلم مصرى مثلى 😅😂
gender wage gap?
You show results for 180 people in a 200 person test.
Where's the results for the other 20 people?
Is this channel run by Tucker Carlson?
EVERYTHING IZ IS CONNECTED. BOLDLY MADE OF THORNS , BOLDLY THOR--NZ IS THE SWEETEST TEA.