Newsletter for the Section on Statistical Education
Volume 9, Number 1 (Winter 2003)
In a brilliant and funny talk given at the end of the ICOTS-6 conference in Cape Town in July last year, George Cobb gave a talk on how to encourage statistical illiteracy in our students. Of course, his aim was to poke fun at unenlightened teaching, uninspired examples, and unclear concepts that make students illiterate, and to provoke us as teachers to teach statistics in ways that prevent this.
It is difficult to summarize the talk and impossible to display the humor that pervaded it, and made it so much more memorable, but I will attempt to summarize his main points, as I heard them.
"If all else fails", he said, "we should lower our standards."
He demonstrated this by giving an example that he had given to a class of his, which related the percentage High School graduation rates in all 50 US States to the egg production of each State, a relationship that, when suitably transformed, is fairly linear. In fact, as egg production increased, high school graduation rates declined.
The point George made was that his students often focused on the technical aspects of the relationship, without realizing, firstly, that the data were not a sample of anything, and secondly, that there was clearly no causal relationship. Nevertheless, students concentrated on identifying p-values, commenting on the meaning of the slope coefficient of the regression line, and speculating on the meaning or otherwise of residuals. George claimed that if he had shown the students this graph before they had taken his introductory statistics class, they would more likely have dismissed the line as irrelevant or meaningless, but after the class, they had been bewitched by formulas and technique to such an extent that they had suspended common sense.
His conclusion is as follows: "[T]he more attention you paid to procedure the less attention you had left for the meaning. That's the single most important insight for creating illiteracy: technique preempts concepts. In a version of Gresham's law, ritual drives out meaning. If you want to drive a wedge between your students and their common sense, just pile on the procedures."
I can highly recommend listening to George give the full version of this talk if you ever have the opportunity to do so. I guarantee that it will bring the house down. I conclude with a quote from the end of the talk: "To review, here are the four secrets again: Variability is everywhere. Bias and confounding are everywhere. Good design is the mother of good data. Good data analysis weaves a tapestry. Conceal these four facts from your students, preferably by teaching lots of procedures and formulas, and you can ensure that they leave your course statistically illiterate."