The Deluge of Spurious Correlations in Big Data
- 1.7k Downloads
Very large databases are a major opportunity for science and data analytics is a remarkable new field of investigation in computer science. The effectiveness of these tools is used to support a “philosophy” against the scientific method as developed throughout history. According to this view, computer-discovered correlations should replace understanding and guide prediction and action. Consequently, there will be no need to give scientific meaning to phenomena, by proposing, say, causal relations, since regularities in very large databases are enough: “with enough data, the numbers speak for themselves”. The “end of science” is proclaimed. Using classical results from ergodic theory, Ramsey theory and algorithmic information theory, we show that this “philosophy” is wrong. For example, we prove that very large databases have to contain arbitrary correlations. These correlations appear only due to the size, not the nature, of data. They can be found in “randomly” generated, large enough databases, which—as we will prove—implies that most correlations are spurious. Too much information tends to behave like very little information. The scientific method can be enriched by computer mining in immense databases, but not replaced by it.
KeywordsBig data Ergodic theory Ramsey theory Algorithmic information theory Correlation
The authors have been supported in part by Marie Curie FP7-PEOPLE-2010-IRSES Grant. Longo’s work is also part of the project “Lois des dieux, des hommes et de la nature”, Institut d’Etudes Avancées, Nantes, France. We thank A. Vulpiani for suggesting the use of Kac’s lemma, G. Tee for providing historical data and A. Abbott, F. Kroon, H. Maurer, J. P. Lewis, C. Mamali, R. Nicolescu, G. Smith, G. Tee, A. Vulpiani and the anonymous referees for useful comments and suggestions.
- Ahn, A. (2015). The party problem. http://mathforum.org/mathimages/index.php/The_Party_Problem_(Ramsey's_Theorem). Accessed December 12, 2015.
- Calude, A. S. (2015). Does big data equal big problems? http://blogs.crikey.com.au/fullysic/2015/11/13/does-big-data-equal-big-problems. November 2015.
- Calude, C. (2002). Information and randomness–An algorithmic perspective (2nd ed.). Berlin: Springer.Google Scholar
- Calude, C. S., & Longo, G. (2015). Classical, quantum and biological randomness as relative. Natural Computing. doi: 10.1007/s11047-015-9533-2
- Cooper, S. B. (2004). Computability theory. London: Chapman Hall/CRC.Google Scholar
- Correlation and prediction. 1992. http://www.intropsych.com/ch01_psychology_and_science/correlation_and_prediction.html
- Devaney, R. L. (2003). An introduction to chaotic dynamical systems (2nd ed.). Redwood City, CA: Addison-Wesley.Google Scholar
- Ferber, R. (1956). Are correlations any guide to predictive value? Journal of the Royal Statistical Society Series C (Applied Statistics), 5(2), 113–121.Google Scholar
- Graham, R. (2007). Some of my favorite problems in Ramsey theory. INTEGERS: The Electronic Journal of Combinatorial Number Theory, 7(2), A2.Google Scholar
- Graham, R., Rothschild, B. L., & Spencer, J. H. (1990). Ramsey theory (2nd ed.). New York: Wiley.Google Scholar
- Grjebine, A. (2015). La dette publique et comment s’en débarrasser. Paris: Press Universitaire de France.Google Scholar
- Grossman, L. (2015). What’s this all about? The massive volume of data that humanity generates is a new kind of problem. The solution is very old: art. Time Magazine, 6 July 2015 (double issue).Google Scholar
- Hoffman, C. (2015). Benchmarked: What’s the best file compression format? http://www.howtogeek.com/200698/benchmarked-whats-the-best-file-compression-format/. May 2015.
- IBM. What is big data? http://www-01.ibm.com/software/data/bigdata/what-is-big-data.html. May 2011.
- Khoussainov, B. (2016). Algorithmically random universal algebras. In M. Burgin & C. S. Calude (Eds.), Information and complexity. World Scientific Series in Information Studies, Singapore, 2016 (to appear).Google Scholar
- Longo, G. (2008). On the relevance of negative results. Influxus. http://www.influxus.eu/article474.html
- Lyons, J. (2013). George Osborne’s favourite “godfathers of austerity” economists admit to making error in research. http://www.mirror.co.uk/news/uk-news/george-osbornes-favourite-economists-reinhart-1838219. April 2013.
- Manin, Y. I. (2016). Cognition and complexity. In M. Burgin & C. S. Calude (Eds.), Information and complexity. World Scientific Series in Information Studies, Singapore, 2016 (to appear).Google Scholar
- Montelle, C. (2011). Chasing shadows: Mathematics, astronomy, and the early history of eclipse reckoning. Baltimore: Johns Hopkins University Press.Google Scholar
- NSF. (2010). Computational and data-enabled science and engineering. http://www.nsf.gov/mps/cds-e/
- Needham, J. (2008). Science and civilisation in China: Medicine (Vol. 6). Cambridge: Cambridge University Press.Google Scholar
- Norvig, P. (2008). All we want are the facts, ma’am. http://norvig.com/fact-check.html
- Oxford Dictionaries. Spurious. http://www.oxforddictionaries.com/definition/learner/spurious. Accessed November 30, 2015.
- O’Grady, C. (2015). Louder vowels won’t get you laid, and other tales of spurious correlation. http://arstechnica.co.uk/science/2015/06/louder-vowels-wont-get-you-laid-and-other-tales-of-spurious-correlation. June 2015.
- Poppelars, J. (2015). OR at work. http://john-poppelaars.blogspot.fr/2015/04/do-numbers-really-speak-for-themselves.html. April 2015.
- Schmidt, E. (2010). Every 2 days we create as much information as we did up to 2003. http://techcrunch.com/2010/08/04/schmidt-data. August 2010.
- Schutt, R., & O’Neil, C. (2014). Doing data science. Newton, MA: O’Reilly Media.Google Scholar
- Sessions, J. (2011). The case for growth: Sessions lists benefits of discretionary cuts. http://www.sessions.senate.gov/public/index.cfm/news-releases?ID=E36C43B4-B428-41A4-A562-475FC16D3793. March 2011.
- Shen, A. (2015). Around Kolmogorov complexity: Basic notions and results. http://dblp.uni-trier.de/rec/bib/journals/corr/Shen15
- Smith, G. (2014). Standard deviations: Flawed assumptions, tortured data, and other ways to lie with statistics. New York: Overlook/Duckworth.Google Scholar
- Smith, J. (2013). From Reinhart and Rogoff’s own data: UK GDP increased fastest when debt-to-GDP ratio was highest—And the debt ratio came down! http://www.primeeconomics.org/articles/1785. April 2013.
- Spurious correlations. http://www.tylervigen.com/spurious-correlations. November 2015.
- Stanton, J. M. (2012). Introduction to data science. Syracuse: Syracuse University.Google Scholar
- Vigen, T. (2015). Spurious correlations. New York: Hachette Books.Google Scholar
- Walkden, C. Magic post-graduate lectures: Magic010 ergodic theory lecture 5. http://www.maths.manchester.ac.uk/~cwalkden/magic/