ASA Connect

Somewhere in my long term memory is a dim recollection of what I thought
of as one of the oldest if not
the oldest examples of a spurious correlation.

What I recall may have come from Kendall & Stewart: it concerns
the relationship between the number of radios licensed in the UK (x) and
the number of people with mental
illness (y), with a strong positive correlation. The punch line, of
course, was that over time, both x and y were
increasing, but for different reasons.

Can anyone supply a reference for this, or an alternative early
reference to an example of spurious correlation?

--
Michael Friendly Email: friendly AT yorku DOT ca
Professor, Psychology Dept. & Chair, Quantitative Methods
York University Voice: 416 736-2100 x66249 Fax: 416 736-5814
4700 Keele Street Web: http://www.datavis.ca
Toronto, ONT M3J 1P3 CANADA

In my teaching regression, I used to quote Mark Twain's Life on the Mississippi (Originally published in 1883, Chapter 17).  So far as I know, this quotation (Attached) is the earliest example of nonsensical correlation.  He wrote the book when he was still Samuel Clemens and taught high school mathematics.

Ajit K. Thakur, Ph.D.
Retired

------------------------------
Ajit Thakur
Associate Director
------------------------------

Original Message:
Sent: 05-16-2017 22:18
From: Michael Friendly
Subject: early examples of spurious correlation

Somewhere in my long term memory is a dim recollection of what I thought
of as one of the oldest if not
the oldest examples of a spurious correlation.

What I recall may have come from Kendall & Stewart: it concerns
the relationship between the number of radios licensed in the UK (x) and
the number of people with mental
illness (y), with a strong positive correlation. The punch line, of
course, was that over time, both x and y were
increasing, but for different reasons.

Can anyone supply a reference for this, or an alternative early
reference to an example of spurious correlation?

--
Michael Friendly Email: friendly AT yorku DOT ca
Professor, Psychology Dept. & Chair, Quantitative Methods
York University Voice: 416 736-2100 x66249 Fax: 416 736-5814
4700 Keele Street Web: http://www.datavis.ca
Toronto, ONT M3J 1P3 CANADA

I cannot give you a reference (I think it was given in an old Penguin book on statistics by Moroney), but an early study showed that in some towns in Germany there was a high correlation between the number of storks and the number of babies born.

------------------------------
Robert Elston
[Emeritus Professor,
Case Western Reserve University]
------------------------------

Original Message:
Sent: 05-17-2017 09:07
From: Ajit Thakur
Subject: early examples of spurious correlation

In my teaching regression, I used to quote Mark Twain's Life on the Mississippi (Originally published in 1883, Chapter 17).  So far as I know, this quotation (Attached) is the earliest example of nonsensical correlation.  He wrote the book when he was still Samuel Clemens and taught high school mathematics.

Ajit K. Thakur, Ph.D.
Retired

------------------------------
Ajit Thakur
Associate Director
------------------------------

Original Message:
Sent: 05-16-2017 22:18
From: Michael Friendly
Subject: early examples of spurious correlation

Somewhere in my long term memory is a dim recollection of what I thought
of as one of the oldest if not
the oldest examples of a spurious correlation.

What I recall may have come from Kendall & Stewart: it concerns
the relationship between the number of radios licensed in the UK (x) and
the number of people with mental
illness (y), with a strong positive correlation. The punch line, of
course, was that over time, both x and y were
increasing, but for different reasons.

Can anyone supply a reference for this, or an alternative early
reference to an example of spurious correlation?

--
Michael Friendly Email: friendly AT yorku DOT ca
Professor, Psychology Dept. & Chair, Quantitative Methods
York University Voice: 416 736-2100 x66249 Fax: 416 736-5814
4700 Keele Street Web: http://www.datavis.ca
Toronto, ONT M3J 1P3 CANADA

Oldenburg, Germany.  See page 8 (Figure 1.4) and its reference in the first (1978) edition of Box, Hunter, Hunter.

------------------------------
Emil M Friedman, PhD
emilfriedman@gmail.com
http://www.statisticalconsulting.org
------------------------------

Original Message:
Sent: 05-17-2017 09:32
From: Robert Elston
Subject: early examples of spurious correlation

I cannot give you a reference (I think it was given in an old Penguin book on statistics by Moroney), but an early study showed that in some towns in Germany there was a high correlation between the number of storks and the number of babies born.

------------------------------
Robert Elston
[Emeritus Professor,
Case Western Reserve University]
------------------------------

Original Message:
Sent: 05-17-2017 09:07
From: Ajit Thakur
Subject: early examples of spurious correlation

In my teaching regression, I used to quote Mark Twain's Life on the Mississippi (Originally published in 1883, Chapter 17).  So far as I know, this quotation (Attached) is the earliest example of nonsensical correlation.  He wrote the book when he was still Samuel Clemens and taught high school mathematics.

Ajit K. Thakur, Ph.D.
Retired

------------------------------
Ajit Thakur
Associate Director
------------------------------

Original Message:
Sent: 05-16-2017 22:18
From: Michael Friendly
Subject: early examples of spurious correlation

Somewhere in my long term memory is a dim recollection of what I thought
of as one of the oldest if not
the oldest examples of a spurious correlation.

What I recall may have come from Kendall & Stewart: it concerns
the relationship between the number of radios licensed in the UK (x) and
the number of people with mental
illness (y), with a strong positive correlation. The punch line, of
course, was that over time, both x and y were
increasing, but for different reasons.

Can anyone supply a reference for this, or an alternative early
reference to an example of spurious correlation?

--
Michael Friendly Email: friendly AT yorku DOT ca
Professor, Psychology Dept. & Chair, Quantitative Methods
York University Voice: 416 736-2100 x66249 Fax: 416 736-5814
4700 Keele Street Web: http://www.datavis.ca
Toronto, ONT M3J 1P3 CANADA

The earliest example I am aware of spurious Pearson's product-moment correlation, is a paper by Pearson himself in 1897, "a Form of Spurious Correlation which may arise when Indices are used in the Measurement of Organs." (paper attached).

In general, it is measuring the magnitude of spurious correlation when ratios of independent measurements are considered. The specific case here are organ/body length ratios of shrimp, reported in an earlier paper by W.F.R Weldon (who comments on the Pearson paper in the same issue).

This is a fascinating early consideration of compositional data analysis, and cited often in the CoDA literature (i.e. by the John Aitchison "school").

------------------------------
Zachary Kurtz
------------------------------

Original Message:
Sent: 05-16-2017 22:18
From: Michael Friendly
Subject: early examples of spurious correlation

Somewhere in my long term memory is a dim recollection of what I thought
of as one of the oldest if not
the oldest examples of a spurious correlation.

What I recall may have come from Kendall & Stewart: it concerns
the relationship between the number of radios licensed in the UK (x) and
the number of people with mental
illness (y), with a strong positive correlation. The punch line, of
course, was that over time, both x and y were
increasing, but for different reasons.

Can anyone supply a reference for this, or an alternative early
reference to an example of spurious correlation?

--
Michael Friendly Email: friendly AT yorku DOT ca
Professor, Psychology Dept. & Chair, Quantitative Methods
York University Voice: 416 736-2100 x66249 Fax: 416 736-5814
4700 Keele Street Web: http://www.datavis.ca
Toronto, ONT M3J 1P3 CANADA

Plot the populations of polar bears vs penguins. 

You can also discuss why storks bring babies. (Think, 9 months before the storks arrive in Germany is???)

------------------------------
Andrew Ekstrom

Statistician, Chemist, HPC Abuser;-)
------------------------------

Original Message:
Sent: 05-16-2017 22:18
From: Michael Friendly
Subject: early examples of spurious correlation

Somewhere in my long term memory is a dim recollection of what I thought
of as one of the oldest if not
the oldest examples of a spurious correlation.

What I recall may have come from Kendall & Stewart: it concerns
the relationship between the number of radios licensed in the UK (x) and
the number of people with mental
illness (y), with a strong positive correlation. The punch line, of
course, was that over time, both x and y were
increasing, but for different reasons.

Can anyone supply a reference for this, or an alternative early
reference to an example of spurious correlation?

--
Michael Friendly Email: friendly AT yorku DOT ca
Professor, Psychology Dept. & Chair, Quantitative Methods
York University Voice: 416 736-2100 x66249 Fax: 416 736-5814
4700 Keele Street Web: http://www.datavis.ca
Toronto, ONT M3J 1P3 CANADA

There's an entire book devoted to that:  Tyler Vigen (2015), "Spurious Correlations."  It's wonderful.  Examples from the book:
Ozone near ground level is 88.7% correlated with functionally obsolete bridges, by year;
Number of Atlantic hurricanes is 79.2% correlated with draft picks by the Boston Celtics;
"Two and a Half Men" ranking against other CBS shows is 81.4% correlated with the NY stock exchange (highest composite price);
Harley-Davidson's motorcycle revenue is 82.1% correlated with the U. S. fertility rate.  Hmm... maybe they have something there...
Gross revenue from U.S. symphony orchestras is 89.2% correlated with juvenile arrests for pot possession.

------------------------------
George Yost
------------------------------

Original Message:
Sent: 05-17-2017 01:12
From: Andrew Ekstrom
Subject: early examples of spurious correlation

Plot the populations of polar bears vs penguins. 

You can also discuss why storks bring babies. (Think, 9 months before the storks arrive in Germany is???)

------------------------------
Andrew Ekstrom

Statistician, Chemist, HPC Abuser;-)
------------------------------

Original Message:
Sent: 05-16-2017 22:18
From: Michael Friendly
Subject: early examples of spurious correlation

Somewhere in my long term memory is a dim recollection of what I thought
of as one of the oldest if not
the oldest examples of a spurious correlation.

What I recall may have come from Kendall & Stewart: it concerns
the relationship between the number of radios licensed in the UK (x) and
the number of people with mental
illness (y), with a strong positive correlation. The punch line, of
course, was that over time, both x and y were
increasing, but for different reasons.

Can anyone supply a reference for this, or an alternative early
reference to an example of spurious correlation?

--
Michael Friendly Email: friendly AT yorku DOT ca
Professor, Psychology Dept. & Chair, Quantitative Methods
York University Voice: 416 736-2100 x66249 Fax: 416 736-5814
4700 Keele Street Web: http://www.datavis.ca
Toronto, ONT M3J 1P3 CANADA

Hello Everyone!

This is not an answer to the original question, but an exemple of a very nice spurious correlation, 'Ice cream and IQ' that appeared in «The Economist». Not only is it an example of a spurious correlation, but it is not even an example of a well-defined correlation since the 2 variables are not (apparently) linearly related. Thus a double example... interesting for teaching purposes!

http://wikistat.mgi.polymtl.ca/tiki-download_file.php?fileId=358

Polymtl		remove preview
View this on Polymtl >

------------------------------
Marc Bourdeau
Ecole Polytechnique
------------------------------

Original Message:
Sent: 05-16-2017 22:18
From: Michael Friendly
Subject: early examples of spurious correlation

Somewhere in my long term memory is a dim recollection of what I thought
of as one of the oldest if not
the oldest examples of a spurious correlation.

What I recall may have come from Kendall & Stewart: it concerns
the relationship between the number of radios licensed in the UK (x) and
the number of people with mental
illness (y), with a strong positive correlation. The punch line, of
course, was that over time, both x and y were
increasing, but for different reasons.

Can anyone supply a reference for this, or an alternative early
reference to an example of spurious correlation?

--
Michael Friendly Email: friendly AT yorku DOT ca
Professor, Psychology Dept. & Chair, Quantitative Methods
York University Voice: 416 736-2100 x66249 Fax: 416 736-5814
4700 Keele Street Web: http://www.datavis.ca
Toronto, ONT M3J 1P3 CANADA

I remember my favorite, an early example I read in some stat book--a correlation of .93 between the annual number of plumbers in the plumbers union in the U.S., and the annual number of inches of rain in the Indian province of Hyderbad (sp) from 1913-1924.

------------------------------
Milton Goldsamt
Consulting Research Statistician
------------------------------

Original Message:
Sent: 05-16-2017 22:18
From: Michael Friendly
Subject: early examples of spurious correlation

Somewhere in my long term memory is a dim recollection of what I thought
of as one of the oldest if not
the oldest examples of a spurious correlation.

What I recall may have come from Kendall & Stewart: it concerns
the relationship between the number of radios licensed in the UK (x) and
the number of people with mental
illness (y), with a strong positive correlation. The punch line, of
course, was that over time, both x and y were
increasing, but for different reasons.

Can anyone supply a reference for this, or an alternative early
reference to an example of spurious correlation?

--
Michael Friendly Email: friendly AT yorku DOT ca
Professor, Psychology Dept. & Chair, Quantitative Methods
York University Voice: 416 736-2100 x66249 Fax: 416 736-5814
4700 Keele Street Web: http://www.datavis.ca
Toronto, ONT M3J 1P3 CANADA

Michael,
You might want to distinguish time-based spurious correlations (the topic of Vigen's book) from "cross-sectional" spurious correlations. A Pearson correlation on almost any non-stationary process is going to be rather large, but that's not what most people care about. The Vigen examples are hilarious (especially the ones on his website), but they are all attributable to the misuse of correlations on time series. More interesting ("significant") are examples that appear to meet the assumptions of computing a bivariate correlation but are due to lurking variables. I'm sure there are some old examples of this sort that people here would know.
Best,
Lee

------------------------------
Leland Wilkinson
H2O
------------------------------

Original Message:
Sent: 05-17-2017 13:41
From: Milton Goldsamt
Subject: early examples of spurious correlation

I remember my favorite, an early example I read in some stat book--a correlation of .93 between the annual number of plumbers in the plumbers union in the U.S., and the annual number of inches of rain in the Indian province of Hyderbad (sp) from 1913-1924.

------------------------------
Milton Goldsamt
Consulting Research Statistician
------------------------------

Original Message:
Sent: 05-16-2017 22:18
From: Michael Friendly
Subject: early examples of spurious correlation

Somewhere in my long term memory is a dim recollection of what I thought
of as one of the oldest if not
the oldest examples of a spurious correlation.

What I recall may have come from Kendall & Stewart: it concerns
the relationship between the number of radios licensed in the UK (x) and
the number of people with mental
illness (y), with a strong positive correlation. The punch line, of
course, was that over time, both x and y were
increasing, but for different reasons.

Can anyone supply a reference for this, or an alternative early
reference to an example of spurious correlation?

--
Michael Friendly Email: friendly AT yorku DOT ca
Professor, Psychology Dept. & Chair, Quantitative Methods
York University Voice: 416 736-2100 x66249 Fax: 416 736-5814
4700 Keele Street Web: http://www.datavis.ca
Toronto, ONT M3J 1P3 CANADA

Original Message------

Michael,
You might want to distinguish time-based spurious correlations (the topic of Vigen's book) from "cross-sectional" spurious correlations. A Pearson correlation on almost any non-stationary process is going to be rather large, but that's not what most people care about. The Vigen examples are hilarious (especially the ones on his website), but they are all attributable to the misuse of correlations on time series. More interesting ("significant") are examples that appear to meet the assumptions of computing a bivariate correlation but are due to lurking variables. I'm sure there are some old examples of this sort that people here would know.
Best,
Lee

------------------------------
Leland Wilkinson
H2O
------------------------------

Original Message------

Michael,
You might want to distinguish time-based spurious correlations (the topic of Vigen's book) from "cross-sectional" spurious correlations. A Pearson correlation on almost any non-stationary process is going to be rather large, but that's not what most people care about. The Vigen examples are hilarious (especially the ones on his website), but they are all attributable to the misuse of correlations on time series. More interesting ("significant") are examples that appear to meet the assumptions of computing a bivariate correlation but are due to lurking variables. I'm sure there are some old examples of this sort that people here would know.
Best,
Lee

------------------------------
Leland Wilkinson
H2O
------------------------------

One of my favorite lurking variable spurious correlations: Children's shoe size and their score on a standardized reading test.

------------------------------
Martha Smith
University of Texas
------------------------------

Original Message:
Sent: 05-17-2017 13:59
From: Leland Wilkinson
Subject: early examples of spurious correlation

Michael,
You might want to distinguish time-based spurious correlations (the topic of Vigen's book) from "cross-sectional" spurious correlations. A Pearson correlation on almost any non-stationary process is going to be rather large, but that's not what most people care about. The Vigen examples are hilarious (especially the ones on his website), but they are all attributable to the misuse of correlations on time series. More interesting ("significant") are examples that appear to meet the assumptions of computing a bivariate correlation but are due to lurking variables. I'm sure there are some old examples of this sort that people here would know.
Best,
Lee

------------------------------
Leland Wilkinson
H2O
------------------------------

Original Message:
Sent: 05-17-2017 13:41
From: Milton Goldsamt
Subject: early examples of spurious correlation

I remember my favorite, an early example I read in some stat book--a correlation of .93 between the annual number of plumbers in the plumbers union in the U.S., and the annual number of inches of rain in the Indian province of Hyderbad (sp) from 1913-1924.

------------------------------
Milton Goldsamt
Consulting Research Statistician
------------------------------

Original Message:
Sent: 05-16-2017 22:18
From: Michael Friendly
Subject: early examples of spurious correlation

Somewhere in my long term memory is a dim recollection of what I thought
of as one of the oldest if not
the oldest examples of a spurious correlation.

What I recall may have come from Kendall & Stewart: it concerns
the relationship between the number of radios licensed in the UK (x) and
the number of people with mental
illness (y), with a strong positive correlation. The punch line, of
course, was that over time, both x and y were
increasing, but for different reasons.

Can anyone supply a reference for this, or an alternative early
reference to an example of spurious correlation?

--
Michael Friendly Email: friendly AT yorku DOT ca
Professor, Psychology Dept. & Chair, Quantitative Methods
York University Voice: 416 736-2100 x66249 Fax: 416 736-5814
4700 Keele Street Web: http://www.datavis.ca
Toronto, ONT M3J 1P3 CANADA

I know someone mentioned  Tyler Vigen's book, but here's a quick look at many on his website (shout out and thanks to my CSP mentee, Naomi Brownstein, who first directed it to me!).
15 Insane Things That Correlate With Each Other



------------------------------
Mary Kwasny
Associate Professor
------------------------------

Original Message:
Sent: 05-16-2017 22:18
From: Michael Friendly
Subject: early examples of spurious correlation

Somewhere in my long term memory is a dim recollection of what I thought
of as one of the oldest if not
the oldest examples of a spurious correlation.

What I recall may have come from Kendall & Stewart: it concerns
the relationship between the number of radios licensed in the UK (x) and
the number of people with mental
illness (y), with a strong positive correlation. The punch line, of
course, was that over time, both x and y were
increasing, but for different reasons.

Can anyone supply a reference for this, or an alternative early
reference to an example of spurious correlation?

--
Michael Friendly Email: friendly AT yorku DOT ca
Professor, Psychology Dept. & Chair, Quantitative Methods
York University Voice: 416 736-2100 x66249 Fax: 416 736-5814
4700 Keele Street Web: http://www.datavis.ca
Toronto, ONT M3J 1P3 CANADA

Original Message------

Somewhere in my long term memory is a dim recollection of what I thought
of as one of the oldest if not
the oldest examples of a spurious correlation.

What I recall may have come from Kendall & Stewart: it concerns
the relationship between the number of radios licensed in the UK (x) and
the number of people with mental
illness (y), with a strong positive correlation. The punch line, of
course, was that over time, both x and y were
increasing, but for different reasons.

Can anyone supply a reference for this, or an alternative early
reference to an example of spurious correlation?

--
Michael Friendly Email: friendly AT yorku DOT ca
Professor, Psychology Dept. & Chair, Quantitative Methods
York University Voice: 416 736-2100 x66249 Fax: 416 736-5814
4700 Keele Street Web: http://www.datavis.ca
Toronto, ONT M3J 1P3 CANADA