What is a spurious correlation?

A spurious correlation is a situation where two variables appear to be related to each other but are actually not, often due to an unseen third factor.

Why does correlation not imply causation?

Because two variables moving together does not mean that one causes the other; there can be other factors at play or the correlation can be coincidental.

What was the real reason behind the decreased heart disease in women taking hormone replacement therapy?

Women in the study had a higher socio-economic status, which included better diet and exercise regimes, and not the therapy itself.

How can small sample sizes affect scientific outcomes?

With smaller sample sizes and many variables, it’s easier to find statistically significant outcomes by chance.

What was the purpose of the chocolate weight loss study?

It was designed to show how science reporting can be manipulated and results misrepresented in media.

What is a lurking variable?

A lurking variable is a hidden factor that may cause both variables to show a correlation without one causing the other.

Why should correlation not be dismissed entirely in science?

Because correlation is often the best available evidence when comprehensive causative studies can't be conducted, and it helps guide further scientific inquiry.

How can scientists avoid being misled by correlations?

By looking for lurking variables, using large and random samples, and validating findings through peer-review or further testing.

This ≠ That

00:03:45

https://www.youtube.com/watch?v=gxSUqr3ouYA

Summary

TLDRThe video highlights the dangers of overvaluing correlations, showing how variables like Nicolas Cage movie releases and swimming pool drownings can misleadingly appear related due to coincidences. It explains that correlation does not imply causation using examples like hormone replacement therapy and nightlight use in children. These examples illustrate 'lurking variables,' where an unseen factor affects both studied variables. The video discloses a deliberately flawed chocolate weight loss study to demonstrate how media may sensationalize science. However, correlation remains vital in scientific research when causative experiments are infeasible. The video encourages critical evaluation of scientific findings and the significance of peer-review processes in research.

Takeaways

📉 Correlation does not imply causation - they can mislead if not analyzed properly.
🧐 Lurking variables can affect both correlated factors without a direct cause-and-effect.
🥼 Proper scientific analysis is crucial for distinguishing real causation from correlation.
🔬 Peer review is essential to validate scientific studies and prevent sensationalism.
⚠️ Be cautious of media representation of scientific results—they can be misleading.
📊 Small sample sizes can produce statistically significant but misleading results.
🍫 The chocolate weight loss study showcased the pitfalls of exploiting correlations.
💡 Correlations still play a critical role in science when causation cannot be directly tested.
📚 Misinterpretation of statistical data is a common issue in popular science reporting.
🔍 Analyzing causative relationships requires meticulous observation and testing.

Timeline

00:00:00 - 00:03:45
The video begins by discussing strange correlations, such as a 66.6% correlation between the number of Nicolas Cage films released and the number of drownings in pools, a 99.26% correlation between divorce rates in Maine and the per capita consumption of margarine, and a 99.79% correlation between spending on Science, Space, and Technology and suicides by specific methods. The video emphasizes that correlation does not imply causation, citing logical fallacies seen in media claims that people who have more sex make more money, without evidence of causal links. It warns that some correlations, like cheese consumption related to bed sheet tangling deaths, offer no real causative relationships.

Mind Map

Video Q&A

What is a spurious correlation?
A spurious correlation is a situation where two variables appear to be related to each other but are actually not, often due to an unseen third factor.
Why does correlation not imply causation?
Because two variables moving together does not mean that one causes the other; there can be other factors at play or the correlation can be coincidental.
What was the real reason behind the decreased heart disease in women taking hormone replacement therapy?
Women in the study had a higher socio-economic status, which included better diet and exercise regimes, and not the therapy itself.
How can small sample sizes affect scientific outcomes?
With smaller sample sizes and many variables, it’s easier to find statistically significant outcomes by chance.
What was the purpose of the chocolate weight loss study?
It was designed to show how science reporting can be manipulated and results misrepresented in media.
What is a lurking variable?
A lurking variable is a hidden factor that may cause both variables to show a correlation without one causing the other.
Why should correlation not be dismissed entirely in science?
Because correlation is often the best available evidence when comprehensive causative studies can't be conducted, and it helps guide further scientific inquiry.
How can scientists avoid being misled by correlations?
By looking for lurking variables, using large and random samples, and validating findings through peer-review or further testing.

View more video summaries

Get instant access to free YouTube video summaries powered by AI!

Subtitles

Auto Scroll:

00:00:00
have you ever wondered how related the
00:00:01
number of people that drown in pools
00:00:03
each year is to how many Nicolas Cage
00:00:05
films are released at the same year
00:00:07
turns out there is a 66.6% correlation
00:00:10
between the two or how about the fact
00:00:11
that there's a ninety-nine point two six
00:00:13
percent correlation between the divorce
00:00:15
rate in Maine and the per capita
00:00:17
consumption of margarine or a ninety
00:00:19
nine point seven nine percent
00:00:20
correlation between spending on Science
00:00:22
Space and Technology and the number of
00:00:25
suicides by hanging strangulation and
00:00:27
suffocation just because there is a
00:00:29
correlation between two variables
00:00:30
doesn't mean that one causes the other
00:00:32
this assumption is a logical fallacy and
00:00:35
yet were drawn to headlines like people
00:00:37
who have more sex make the most money
00:00:39
unfortunately this does not always equal
00:00:41
that even though it may look like it
00:00:44
cheese consumption probably isn't
00:00:46
related to how many people died tangled
00:00:48
in their bed sheets but sometimes it
00:00:50
isn't so obvious numerous studies found
00:00:52
that menopausal woman taking hormone
00:00:54
replacement therapy had a lower than
00:00:56
average incidence of heart disease
00:00:57
leading doctors to believe that hormone
00:00:59
replacement could protect against heart
00:01:01
disease however when women underwent
00:01:03
randomized control trials they found
00:01:05
that hormone replacement therapy
00:01:07
actually increased the risk of heart
00:01:09
disease when the original data was
00:01:10
finally reanalyzed it was found that
00:01:12
women who took the therapy were of a
00:01:14
higher socio-economic group with a
00:01:16
better diet and exercise regime this was
00:01:18
the real cause behind decreased risk of
00:01:20
heart disease
00:01:21
another case found that those who used
00:01:23
nightlights as a kid were more likely to
00:01:25
develop myopia but there's actually a
00:01:27
strong link between parental myopia and
00:01:29
the development of child myopia so in
00:01:31
reality myopic parents were simply more
00:01:33
likely to leave a light on in their
00:01:35
child's bedroom these are examples of a
00:01:37
lurking variable where a does not cause
00:01:39
B but rather C causes them both it's a
00:01:42
little like taking people who have lung
00:01:43
cancer and thinking hey they're all
00:01:46
carrying lighters in their pockets so
00:01:48
lighters must cause cancer while not
00:01:50
realizing that smoking is the
00:01:51
confounding variable and scientists work
00:01:54
hard in their studies to try and avoid
00:01:55
this but it gets worse when popular
00:01:57
media takes advantage of these
00:01:59
potentially coincidental correlations
00:02:01
and one study about the chocolate weight
00:02:03
loss connection was actually designed as
00:02:05
a way to expose how science reporting
00:02:06
can be
00:02:07
sation alized a science writer with a
00:02:09
PhD in microbiology ran a real clinical
00:02:11
trial where participants were assigned
00:02:13
to three groups a low-carb diet a
00:02:15
low-carb diet plus a one point five
00:02:17
ounce
00:02:17
chocolate bar and a group that maintains
00:02:19
their regular diet at the end of three
00:02:21
weeks the chocolate group did lose the
00:02:23
most weight but the journalist
00:02:25
consciously used terrible science he
00:02:27
used 15 participants and measured 18
00:02:29
different measurements including weight
00:02:31
loss cholesterol sleep quality blood
00:02:33
pressure well-being etc and when you use
00:02:36
a small group of people and measure a
00:02:37
large number of things you're pretty
00:02:39
much guaranteed to get a statistically
00:02:41
significant result which veritasium has
00:02:44
an amazing video on here if you want to
00:02:46
check out the result could have easily
00:02:47
been something different such as
00:02:49
chocolate correlates to lower blood
00:02:50
pressure if this study had been peer
00:02:52
reviewed by other researchers it would
00:02:54
have been called out so instead he
00:02:56
submitted it to a journal for a fee of
00:02:57
600 euros making up a fake institution
00:03:00
named the Institute of diet and health
00:03:02
he then sent out a press release to
00:03:04
dozens of media publications and very
00:03:07
quickly getting slim buy chocolate was
00:03:09
front-page news with that in mind
00:03:11
however we can't dismiss correlation
00:03:13
entirely correlative evidence is an
00:03:15
essential part of science double-blind
00:03:17
studies are not always possible or
00:03:19
ethical to run often leaving correlation
00:03:21
as the best evidence available when
00:03:23
every possible causative relationship is
00:03:25
systematically explored correlation can
00:03:27
be used as a powerful tool for assessing
00:03:29
cause-and-effect relationships and
00:03:30
progressing science even further
00:03:32
big thanks to Tyler vegan for providing
00:03:34
his charts on these interesting
00:03:35
correlations you can check out his
00:03:36
website or his book spurious
00:03:38
correlations for more peculiar examples
00:03:40
and subscribe for more weekly science
00:03:42
videos every Thursday