Month: July 2015

My publications are causing global warming: spurious correlations with my academic performance

Posted on Updated on

Source: xkcd.com
Source: xkcd.com

Recently, California governor Jerry Brown signed a bill exiling personal and religious beliefs as legitimate exemptions for child vaccinations. While I personally commend Gov. Brown for taking such a strong stance in the interest of public health, many do not share my optimism. Unsurprisingly, proponents of the anti-vaccination movement (like Jim Carrey) spoke out against the decision. As you likely know, these individuals protest vaccinations on the assumption that chemical preservatives in the vaccines cause autism, with many “anti-vaxxers” often stating that their son or daughter showed signs of autism only after getting their shots. These claims of vaccinations causing autism stem from a debunked and retracted scientific article suggesting that an increase in autism diagnoses was linked to the appearance of the MMR vaccination. Since the paper’s retraction, numerous studies have failed to find any link between vaccines and autism. Furthermore, not only was this paper retracted because the data were falsified, but the causal conclusions of the paper were based on correlational evidence, much like the statements of anti-vaxxers with autistic children – a logical fallacy in the realm of good science.

Autism prevalence over time in California (without open points) and the UK (with open points) fitted with introduction of MMR vaccine (arrows). This graph is often used to support the claim that vaccines cause autism. Not only were these data fabricated, but such a link has been refuted by numerous studies since.
Autism prevalence over time in California (with open points) and the UK (without open points) fitted with introduction of MMR vaccine (arrows). This graph is often used to support the claim that vaccines cause autism. Not only were these data fabricated, but such a link has been refuted by numerous studies since. Source: Wakefield et al. 1999. The Lancet 354:949-950.

Although correlation does not imply causation, correlations can often provide hints of a causal relationship. However, when simply presented with a correlation, four outcomes are possible:

  1. A causes B
  2. B causes A
  3. A and B are related by a common causal agent
  4. A and B are completely unrelated but have coincidentally similar patterns/trends

As such, it is important to look at correlational evidence with a critical eye and explore mechanisms of causality that lend evidence for a causal relationship before making any substantial claims. For example, the correlation between increasing atmospheric CO2 concentrations, increasing global temperature, and anthropogenic CO2 emissions would not be sufficient to conclude with confidence that increasing anthropogenic CO2 is causing global warming. However, a plethora of studies reporting that increasing CO2 in the atmosphere really does lead to increasing temperature, that other sources of atmospheric CO2 are negligible in comparison to anthropogenic sources, and that other pollutants do not contribute to increasing temperature as much as CO2 does do provide evidence that increasing atmospheric CO2 is caused by human activity, and that this is the cause of global warming. Although such exploration may uncover a causal link, many correlations simply fall under #4 above. To illustrate such non-causal correlations, various articles have been published online. Most notably, a website – Spurious Correlations – highlights bizarre and often hilarious non-causal correlations.

To join in on the spurious correlation fad, I decided to look at some strange correlations with my academic performance thus far by comparing the number of papers I’ve published during each year of my Ph.D. between 2012 and 2015 (2015 data reflect papers that have been published, submitted, are in press, or are in prep); 10 such correlations are below. All of these correlations are strictly non-causal and the relationships are not linked in any way – they are simply meant for your entertainment by imagining the comedic nature of causal links between the variables.

1. Annual publications correlate with atmospheric CO2

Correlation: 99%

Clearly my publications are causing global warming, right “skeptics”?

pubs v CO2

2. Annual publications correlate with McDonald’s net income

Correlation: 89%

Ba-da-ba-ba-ba, I’m publishing!

pubs v mcdonalds

3. Annual publications correlate with cumulative annual spending on advertisements by mobile companies

Correlation: 99%

Maybe smartphones really do make you smarter.

pubs v mobileads

4. Annual publications correlate with ratings of “Suits” finales

Correlation: 99%

Maybe my co-authors have lost interest in Harvey Specter.

pubs v suitsratings

5. Annual publications correlate with Alexander Ovechkin’s scoring productivity

Correlation: 84%

Should I thank Ovi, or should Ovi thank me?

pubs v ovechkingoals

6. Annual publications correlate with the legalization of same-sex marriage in the USA

Correlation: 98%

Seems like America and I are both doing something right.

pubs v samesexmarriage

7. Annual publications correlate with Nicolas Cage movie appearances

Correlation: 98%

It’s not exactly mai-thais and yatzee out here.

 pubs v cagefilms

8.Annual publications correlate with airplane accidents

Correlation: 89%

If it was paper airplanes, this might actually be causal.

pubs v planecrashes

9. Annual publications correlate with US chain store sales

Correlation: 92%

I don’t think journal fees are included in chain store sales.

pubs v chainsales

10. Annual publications correlate with the price of coffee

Correlation: 95%

And the winner for “correlation most likely to be causal” goes to…

pubs v coffeesales

Advertisements