Spurious Connections


Students in introductory psychology classes often struggle with correlations. Even if you have them stand on their desks ala Dead Poets Society and recite “Correlations do not equal causation” many continue to misinterpret correlational research. 

At Soomo, we have developed an Investigation, Correlation is Not Causation, to teach the students how to evaluate reported conclusions. However, we all know that students can use still more help with this topic.

In my own teaching, showing students how correlational data is gathered was a big step in helping them understand the kind of conclusions that can be made. My favorite resource for this is Spurious Correlations, a simple website that presents the correlations between random variables drawn from a number of publicly available sources. My students never failed to laugh at the moderately positive correlation between “Films Nicolas Cage appeared in” and “Number of people who drown by falling into a pool.”

The website’s creator, Tyler Vigen, started it as a “fun way to look at correlations and to think about data.” And it has been very effective to help my students do just that. These spurious correlations drive home the point that variables may be statistically related even if there is no logical relationship.Pasted Graphic 1

How to Incorporate the Website into Your Class:

Discuss data collection.   Pick one of the spurious correlations and ask students if the data could be obtained through an experimental manipulation of variables. For example, do they think someone actually conducted an experiment to determine if increasing the Age of Miss America would increase the number of murders per year by “steam, hot vapours and hot objects”? If no experiment was conducted, then how was the data obtained? You can then share the methods Mr. Vigen uses, simply finding variables in a one or more existing data sets, and contrast with an experiment that manipulates variables. If your students seem up for it, you could extend this conversation to discuss how Mr. Vigen’s method does or does not follow the scientific method of research. You might note that he does not start with a hypothesis then carefully choose data to test it, but instead gathers data and searches for interesting correlations. To our delight, of course.

Students pick a Spurious Correlation to examine.  A great in-class or online discussion board activity would be to assign - or ask students to choose – one of the Spurious Correlations (there are over 30,000!) and answer the following questions as an individual or small group:

  • Describe the direction of the correlation; is it positive or negative? Describe the strength of the correlation; is it weak, moderate, or strong? Students may need some direction to determine the strength of correlational coefficients. This image is helpful: http://image.mathcaptain.com/cms/images/41/pearson-correlation-coefficient-interpretation.jpeg 
  • Can you draw a causal or a relational conclusion from this data? 
  • What does this correlational data mean? Is it useful to understand behavior or is it a random finding? 
  • Describe an experiment that could test whether systematically manipulating Variable X could cause a change in Variable Y (if you were not constrained by ethics, methods, money, etc.).

Conduct a Correlation Together.  A new feature of the site is “Discover a correlation” which quickly presents the coefficients and graphs of the correlation between two variables from a number of interesting data sets. Walk through this process with your students in class or in the form of a video, asking aloud if there is a reason to assume a connection between the two variables.

  • For example, ask students to develop a hypothesis for the relationship between “Precipitation in North Carolina” and “Precipitation in South Carolina.” Do they think there will be a positive or negative correlation? A weak or strong correlation?
  • Discuss the resulting graph, the coefficient, and the conclusion. Was the hypothesis supported?
  • After a couple of guided demonstrations, have students conduct two of their own tests. Ask them to first choose variables that they believe are logically related and then choose variables that are likely “spurious.” Ask them to share in a presentation or discussion board the resulting graphs, the coefficient, and their conclusions along with a reflection of what they have learned about correlational research from this activity.

We would love to hear how you teach students about correlational research!

by jennifer in research