The 'correlation is not causation' hockey stick

http://books.google.com/ngrams/graph?content=correlation+is+not+causation&year_start=1800&year_end=2008&corpus=0&smoothing=3

Mike Lorrey writes- PAY ATTENTION CLIMATE ALARMISTS:

“The phrase ‘correlation does not imply causation’ goes back to 1880 (according to Google Books). However, use of the phrase took off in the 1990s and 2000s, and is becoming a quick way to short-circuit certain kinds of arguments.

In the late 19th century, British statistician Karl Pearson introduced a powerful idea in math: that a relationship between two variables could be characterized according to its strength and expressed in numbers. An exciting concept, but it raised a new issue: how to interpret the data in a way that is helpful, rather than misleading. When we mistake correlation for causation, we find a cause that isn’t there, which is a problem. However, as science grows more powerful and government more technocratic, the stakes of correlation — of counterfeit relationships and bogus findings — grow larger.”

From Slashdot: The History of ‘Correlation Does Not Imply Causation’

==============================================================

From the Slate article referenced by Slashdot:

The graph below, again from Google Books, shows the shift in language that marked this change in spirit: Up until the early 1900s, causation showed up more often than correlation in the corpus; then the concepts flip. (I’ll let someone else explain why correlations have been trending downward since 1976.)

http://books.google.com/ngrams/graph?content=correlation%2C+causation&year_start=1800&year_end=2010&corpus=0&smoothing=3

 

Advertisements

  Subscribe  
newest oldest most voted
Notify of
corio37

When I studied Psychology twenty years ago it was anxious to be regarded as a ‘real’ science, and so we got an excellent grounding in scientific method. Amongst other things we were told never to use the word ’causes’ in a scientific paper where ‘correlates with’ was correct. Unfortunately it looks like some Psychology faculties have forgotten that useful lesson.

cbb

An old professor said to us undergraduates, “Boiled lobsters are red. Not everything that is red is a boiled lobster”.

anna v

The graph only supports that the term “correlation” became popular . Causation seems pretty constant, it did not pick up the slack after 1976. It is probably fashions in scientific and semi language.

anna v

that is “semi-scientific” in the last line.

Gunga Din

Every day I that I go to work when I come home from work, my dog wants a treat.
Therefore, if I never went to work, my dog would never want a treat.
Or maybe if I never came home from work, my dog would never want a treat?
More reasearch funds are needed.

From Graph #1….”Correlation is not causation” is trending upward….
From Graph #2….”Causation” peaked in 1885…oscillatted briefly….trended downward in 2000….
“Correlation” overtook in 1900…peaked in 1976….rapid decline since….
Conclusion: We are approaching the happy 1800 era when there was NO causation and NO correlation….Ignorance is Bliss….Orwell would be happy….

Thomas T

The better term would be, “correlation does not prove causation”, because correlation can damn well be a strong enough clue to be suggestive of causation. CO2 and global temperature is a case in point. Specifically, it was proposed over 100 years ago that the contribution of CO2 rise to temperature increase can be estimated. The subsequent observed correlation of CO2 and global temperature is thus suggestive, but not proof, of causation.
What is important here is that at present no other parameter, such as solar or earth tilt, etc, correlates well with the observed rapid increase in global temperature.
That is the kicker, and so, although correlation does not prove causation:
If you hear hoofbeats, by all means keep looking for zebras, but smart money will be that you hear a horse.

D Böehm

Thomas T,
Yes, correlation is not causation. But when you look at this, you are forced to admit that the only correlation between CO2 and temperature is that temperature changes cause CO2 changes; not vice-versa. There is no empirical scientific evidence showing that changes in CO2 cause changes in temperature. Draw your own conclusions.

RoHa

@ corio37
“When I studied Psychology twenty years ago it was anxious to be regarded as a ‘real’ science, and so we got an excellent grounding in scientific method.”
Same when I studied it 45 years ago. That grounding in scientific method added to the grounding that high school science had given me, and included a fair chunk of statistics. I learnt two very important things.
1. Aspirations notwithstanding, psychology wasn’t a science.
2. Get a professional statistician to do the stats for you.

0U812

Oh Dear. A ‘Mann’ made Éruption Volcanique thus looms large and when ignited will issue for many weeks to come.
Tinfoil hat firmly attached, lead fused lab coat on, and now in the Level-B20 Lead-Lined bunker.
[Via the Radio Dispatch] “I have shorted the locks. The doors are now sealed. I’ll remain here and keep recording the events for as long as possible.
Later, you will eventually be able to retrieve all the tapes from the vault.
For now I wish you, God Speed. Make Haste as you have little time.
As you exit the research facility through the hills look to the east. A New Moon is about to be born and you will witness that event.
As for me …. I did all that I could; History will show that. Signing Off.” [/Via the Radio Dispatch]
XD

David Ball

It appears the time series selected is differing between graphs.

Thomas T

D. Boehm, consider feedback mechanisms before you force yourself to conclude. If you do so, then your pressure on yourself will evaporate.

theduke

Causation is much more difficult to prove. Everything is related. Similarities abound. Trends and characteristics overlap. Slopes can be equal. Correlation is everywhere. Lobsters are red and so are apples (unless they are green and then they correlate with grass.) Apples do not grow in the ocean, but to the untrained eye, they correlate with boiled lobsters. It’s arguable that they correlate because the similar color results from exposure to heat, but . . . .Oh, never mind.
Repeat: Causation is much more difficult to prove.
Forgive me, but the conclusion that increases in the trace gas CO2 are responsible for the slight increase in heat we’ve experienced over the past 100 years because establishment climate scientists refuse to vigorously explore (and disprove) other possible causes should be offensive to anyone who respects the scientific method.

D Böehm

Thomas T,
Positive feedbacks have been falsified by Planet Earth; the ultimate Authority. Believe what you want, it isn’t reality. CO2=CAGW is a debunked fantasy.

Mooloo

What is important here is that at present no other parameter, such as solar or earth tilt, etc, correlates well with the observed rapid increase in global temperature.
What temperature causes CO2 increases? The whole deck of cards comes tumbling down.
Correlation should not even imply causation. Especially if your “proof” is the “I can’t think of anything else” type argument.
At the best it is, as I tell my students, “consistent with”.

Thomas T:
So if Variable 1 and Variable 2 are correlated, and change in Variable 1 occurs before change in Variable 2, what can you conclude?
(Hint: Variable 1 is NOT CO2 concentraion, Variable 2 is NOT Temperature.)

theduke

D. Boehm: but CO2=.10XAGW (or something like) is still a possibility.

Gunga Din

Thomas T says:
October 2, 2012 at 8:53 pm
The better term would be, “correlation does not prove causation”, because correlation can damn well be a strong enough clue to be suggestive of causation. CO2 and global temperature is a case in point. Specifically, it was proposed over 100 years ago that the contribution of CO2 rise to temperature increase can be estimated. The subsequent observed correlation of CO2 and global temperature is thus suggestive, but not proof, of causation.
What is important here is that at present no other parameter, such as solar or earth tilt, etc, correlates well with the observed rapid increase in global temperature.
That is the kicker, and so, although correlation does not prove causation:
If you hear hoofbeats, by all means keep looking for zebras, but smart money will be that you hear a horse.
==============================================================
Whether you probably hear a zebra or a horse or a bison depends on where you live or where you are in the zoo.
“What is important here is that at present no other parameter, such as solar or earth tilt, etc, correlates well with the observed rapid increase in global temperature.”
Really? “Rapid increase in global temperature”? Where were these measuring devices sited? Has anything happened since they were set that might bias there records? Has any bias entered in that has influenced what numbers have actually been enter and/or altered in the records? Temperatures don’t correlate with what the Sun has been up to? How about El Nino and La Nina? (Check out Tisdale’s work.)
PS Please define just how “rapid” and just how much of an “increase” has occured and where it occured and over what period of time.
PSPS Are you a member of the BBSB?

Go Home

“If the Washington Redskins win their last home game before the election, the incumbent party retains the White House. Otherwise, the out-of-power party wins.
This pattern has failed only once since 1940, though which year it failed is something of a debate on Wikipedia, depending whether one honors a retroactive change to the rule to recognize the winner of the popular vote.”
http://news.yahoo.com/your-nfl-team-can-predict-the-election.html
Sounds like warmists arguing correlation implying causation…
++++++++++++++++
If Denver rushes for more than 106 yards in its fifth game of the season, the incumbent party wins. Otherwise, the out-of-power party wins. This rule has held for every election since the Denver Broncos’ first season in 1960.
Here is the Patriots Rule, true for 12 of the past 13 elections: If New England commits fewer turnovers than its opponent in the team’s first away game, the out-of-power party will win the White House. Otherwise, the incumbent party wins.
{ONE FOR EACH TEAM)
Saving the cardiac cards for last… the Cardinals Rule: If Arizona averages more than 18.2 yards per first down in its fifth-to-last game before the election, the out-of-power party wins. Otherwise, the incumbent party retains the White House. {17 out of 17)
This Thursday: at St. Louis on Oct. 4
GO CARDS or GO HOME

“One of the largest-ever studies of retractions has found that two-thirds of retracted life-sciences papers were stricken from the scientific record because of misconduct such as fraud or suspected fraud — and that journals sometimes soft-pedal the reason. The study contradicts the conventional view that most retractions of papers in scientific journals are triggered by unintentional errors. The survey examined all 2,047 articles in the PubMed database that had been marked as retracted by 3 May this year. But rather than taking journals’ retraction notices at face value, as previous analyses have done, the study used secondary sources to pin down the reasons for retraction if the notices were incomplete or vague. The analysis revealed that fraud or suspected fraud was responsible for 43% of the retractions. Other types of misconduct — duplicate publication and plagiarism — accounted for 14% and 10% of retractions, respectively. Only 21% of the papers were retracted because of error (abstract).”
http://science.slashdot.org/story/12/10/02/180226/misconduct-not-error-is-the-main-cause-of-scientific-retractions

theduke

I wrote: D. Boehm: but CO2=.10XAGW (or something like) is still a possibility.
Make that CO2=.10xGW

Barry K

Thomas T,
Before humans became a factor, temperature variations an order of magnitude larger than what we’ve seen in the recent century occurred. Until scientists develop a model that can generate such changes without human influence (i.e. CO2), it’s clear that the models are neglecting some unknown effects….

Thomas T

establishment climate scientists refuse to vigorously explore (and disprove) other possible causes
Name an establishment scientist that has refused to explore and disprove other possible causes. And once you name them, what is your evidence?

davidmhoffer

Oh gimme a break.
Pick up a book that was written in the early 1800’s. You’ll find words in it that were common then and all but abandoned now. You’ll find common phrases that are no longer in use. You’ll find styles of expression that were the way regular folks talked that seem quaint or odd or stilted today. Anyone get to the saloon for a tipple lately? Anyone in 1820 mention renting a video and picking up some microwave popcorn?
If “link” or “connection” or “association” or any number of other words were commonly used instead of “correlation” at various points in time, you’d esentially be getting a false negative because they were excluded from the search. Were they? I haven’t a clue.
But language evolves over time and any study of terminology use that ignores that fact is just meaningless.

Robin Melville

I was warned about this fallacy by my Stats teacher who pointed out that when temporal and spatial measurements were taken into account it’s trivial to prove that, in England, churches cause pubs.

Go Home

Continuing with the “NFL predicts presidential election theme” above…
So you put all team predictors into a climate model with each having excellent correlation factors (only 7 failures out of 32 predictors at approximately 10 samples per). This thing should be fail proof. But can it predict this years election?
“Four rules were decided in the past week. The Bears Rule and the Titans Rule point to an Obama win, while the Bills Rule and the Buccaneers Rule point to a Romney win. The score is now 9-4 in Mitt Romney’s favor.”
Already four to nine failures out of 13 predictions. When the last 65 years had a total of only 7 failures. Whats going wrong this year?

For the umpteenth time, absent chance (coincidence), correlation is proof of causation. It’s just that correlation between A and B isn’t proof that A causes B, which is the fallacy.
Wikipedia has a good summary,
The cum hoc ergo propter hoc logical (correlation proves causation) fallacy can be expressed as follows:
A occurs in correlation with B.
Therefore, A causes B.
In this type of logical fallacy, one makes a premature conclusion about causality after observing only a correlation between two or more factors. Generally, if one factor (A) is observed to only be correlated with another factor (B), it is sometimes taken for granted that A is causing B, even when no evidence supports it. This is a logical fallacy because there are at least five possibilities:
A may be the cause of B.
B may be the cause of A.
some unknown third factor C may actually be the cause of both A and B.
there may be a combination of the above three relationships. For example, B may be the cause of A at the same time as A is the cause of B (contradicting that the only relationship between A and B is that A causes B). This describes a self-reinforcing system.
the “relationship” is a coincidence or so complex or indirect that it is more effectively called a coincidence (i.e. two events occurring at the same time that have no direct relationship to each other besides the fact that they are occurring at the same time). A larger sample size helps to reduce the chance of a coincidence, unless there is a systematic error in the experiment.
http://en.wikipedia.org/wiki/Correlation_does_not_imply_causation

Breaker

So, is the point of the graph that the use of the word “correlation” has caused global warming and, with a five year or so lag, the cooling trend of the 2000’s? I guess Ronald Reagan’s saying, “Words have Consequences,” was more accurate than we thought.

JJ

Thomas T says:
The better term would be, “correlation does not prove causation”, because correlation can damn well be a strong enough clue to be suggestive of causation.

No. “Correlation does not prove causation” is a fine alternative wording, but the strength of a correlation does not have a damn thing to do with whether or not causation is also involved.
What is important here is that at present no other parameter, such as solar or earth tilt, etc, correlates well with the observed rapid increase in global temperature.
No. What is important here is that right under your “Correlation does not prove causation” sign, you need to pencil in “And ignorance doesn’t prove anything.”
That is the kicker,…
It is not a kicker. It is a logical fallacy. A variant of argumentum ad ignorantiam.

AndyG55

@theduke “Correlation is everywhere. Lobsters are red and so are apples (unless they are green and then they correlate with grass.) Apples do not grow in the ocean, but to the untrained eye, they correlate with boiled lobsters”
You really don’t understand the meaning of “correlation” do you. !!!

AndyG55

@Phillip “( two events occurring at the same time that have no direct relationship to each other besides the fact that they are occurring at the same time)”
And in the case of CO2 and global average (urban) land temperature, this ONLY happened for a period between 1976ish and 1998ish…
which ALSO happened to coincide with massive urbanisation in many areas where land thermometers were located.
Now that they are being more careful with land temp calcs (because they are being watched) and satellite temps are now accepted, the coincidence of CO2 and temperature increases has DISAPPEARED !!

theduke

Thomas T says:
October 2, 2012 at 10:03 pm
establishment climate scientists refuse to vigorously explore (and disprove) other possible causes
Name an establishment scientist that has refused to explore and disprove other possible causes. And once you name them, what is your evidence?
——————————————————————-
All of them. Cite a paper where someone associated with the IPCC has attempted to explore other possible causes for global warming than man-made CO2. Better yet, cite a paper that questions the orthodoxy that CO2 is the cause of AGW. Cite a paper that proves that causation and not correlation validates the theory of CO2 induced AGW.
They decided a priori a long time ago that CO2 was the cause based on extrapolating a century-old theory. They haven’t bothered to prove it, because they’ve gotten away with relying on correlation.
Junk science.

theduke

Thomas T wrote: “The better term would be, “correlation does not prove causation”, because correlation can damn well be a strong enough clue to be suggestive of causation.”
So you want governments around the world to spend (or cause people to spend) trillions of dollars because you believe you have a “strong enough clue to be suggestive of causation?”
That sounds “damn well” stupid to me.

AndyG55

Robin Melville says:
“I was warned about this fallacy by my Stats teacher who pointed out that when temporal and spatial measurements were taken into account it’s trivial to prove that, in England, churches cause pubs.”
Robin, you have the causality wrong way around 😉
Churches are built, to cope with the evil of pubs……… or maybe…
hmmmm this requires some more research.. …… funding please someone…
I’ll do the pub part of the research.

Doug UK

Gunga Din says:
October 2, 2012 at 8:49 pm
Every day I that I go to work when I come home from work, my dog wants a treat.
Therefore, if I never went to work, my dog would never want a treat.
Or maybe if I never came home from work, my dog would never want a treat?
More reasearch funds are needed.
………………
And if you prefaced your research grant request with “The effect of Climate Change on Canine Pavlovian responses” – the money would be yours.

DonK31

Simple logic tables:
If A then B.
If B…what about A?
A may or may not be true.

Reblogged this on Standard Climate.

Let’s remember Richard Feynman: http://www.youtube.com/watch?v=b240PGCMwV0
Correlation is a statistical way to determine if experimental data fit with a theoretical model, and to compute a index of confidence for it.
But a theoretical model canot be deducted from a correlation:
– In the Northern Hemisphere chimneys are beginning to make more smoke in the Autumn.
– At this same time leaves are falling from the trees.
– There is a correlation between these two observations.
– Therefore: is smoke causing the falling of leaves?
Too many people make such mistakes because they don’t analyze all consequences implied by the theory that they expose.

richardscourtney

Philip Bradley:
At October 2, 2012 at 10:29 pm you write

For the umpteenth time, absent chance (coincidence), correlation is proof of causation. It’s just that correlation between A and B isn’t proof that A causes B, which is the fallacy.

For the umpteenth time YOU ARE WRONG . Correlation proves nothing.
Any two time-series will correlate for some periods of the time if they are long enough.
Richard

Urederra

Lack or correlation proves lack of causation.
That is what we should apply to CO2 and temperatures.
I remind you people that Pearson was talking about quantified variables, or so I understood.

Urederra

Sorry, It should read. “Lack of correlation proves lack of causation”

dave38

Robin Melville says:
October 2, 2012 at 10:08 pm
I was warned about this fallacy by my Stats teacher who pointed out that when temporal and spatial measurements were taken into account it’s trivial to prove that, in England, churches cause pubs.
The reason is actually very simple.
The Bible tells us to thirst after righteousness!

peterk505

Plot a graph of ice-cream sales over a year.
Plot a graph of swimming pool drownings over a year.
Ah, we seem to have a correlation- they both peak around August!
Now the burning question is- do ice-cream sales affect swimming pool drownings or vice-versa?
Hmmm……

As I’ve argued before, for the purposes of the scientific side of the climate debate, the “correlation does not necessarily imply causation” element of the fallacy is the most important one. Not only is there no empirical evidence to support the contention that anthropogenic GHG emissions control global temperatures; the graphs of the two phenomena don’t even line up. So I don’t understand why people are even looking for evidence (yes, that was a rhetorical statement; I understand why people are looking for evidence of a causal linkage between delta T and delta CO2 – but the reasons have nothing to do with science).
However, from the policy side of the climate debate, the inverse argument is more important, i.e., that non-correlation demonstrates non-causation, or in other words, if there is no statistically significant correlation between two phenomena, it is impossible to posit a causal relationship between them. The IPCC itself admits this by arguing that “temperature response” should “scale linearly with forcings”, of which the most important (in their opinion) is human GHG emissions. The crux of the debate, therefore, should hinge on the fact that there is no statistically significant correlation between change in CO2 concentration and change in temperature over any time period for which reliable temperature records and/or proxies exist – with (again, as I’ve mentioned before) the sole example of Antarctic ice core samples, which seem to demonstrate a correlation, but with delta T preceding delta CO2, thus implying that the former causes the latter rather than the inverse.
Moreover, and again from a policy perspective, if there is no statistically significant correlation between delta T and delta CO2, how can you design a policy to limit change in the former by constraining anthropogenic emissions of the latter? This basic question seems to keep getting lost in the manufactured (and totally irrelevant) furor over tenths of degrees.

Dale

I get older by 1 per year. CO2 goes up by 2 ppm per year.
Therefore my age determines CO2 ppm.
In 1974, CO2 was ~330 ppm. Thus, extrapolating out to when I’m 60 (2034), CO2 will be 450 ppm.
Correlation == causation! It’s worse than we thought since CO2 – Temps is another “correlation == causation”, that means temps will be ~3.2C higher than pre-Industrial in 2034!
BS!

David S

Philip Bradley
Absent chance…or correlation the other way…or another causative agent, or any mixture of the three, ie absent everything else, so argumentum ad ignorantem in other words.

AndyG55

“Any two time-series will correlate for some periods of the time if they are long enough.”
And in the case of CO2 and temperature, a VERY SHORT 20 odd year period, with NO CORRELATION (negative in fact) between 1940-1970ish OR between 1998 and now.
so over the time that there has been any significant human CO2 release, the correlation has only been for about 1/3 of the time…. and these MORONS try to attribute causation.. DOH !!!!

AndyG55

sorry, I used “temperature”.. where I really meant “homogemnised NH Urban land temperature”

H.R.

Gunga Din says:
October 2, 2012 at 8:49 pm
“Every day I that I go to work when I come home from work, my dog wants a treat.
Therefore, if I never went to work, my dog would never want a treat.
Or maybe if I never came home from work, my dog would never want a treat?
More reasearch funds are needed.”

Fess up; your dog has you well-trained ;o)