Correlation
391
The Meaning of Correlation
The Pearson Correlation Coefficient
Applying Correlation in Psychometrics
Next Steps: Partial Correlation
392
John Snow’s map of the London cholera epidemic (Chapter 1) revealed a correlation, a systematic association (or relation) between two variables. Researchers from the Massachusetts Institute of Technology (MIT) used a correlation to reveal the dangers of academic cheating (Palazzo, Lee, Warnakulasooriya, & Pritchard, 2010). They compared the final exam scores of 428 physics students with how frequently they had copied 3 days of computerized homework assignments from classmates. They assumed that the minority of students (< 10%) who finished their homework within a ridiculously short time had cheated. You might expect cheating to boost grades—
What might be going on here? Students’ preexisting ability in math and physics was not correlated with cheating, so these were not weak students who were also incompetent at cheating.
Their computerized homework revealed a correlation that suggested another possible answer: Cheaters procrastinated by starting their homework too late to complete it without cheating. Did they have outside jobs, suffer from anxiety, or have other commitments that cut into homework time? Correlations can’t tell us which explanation is right, but they can force us to think about the possible explanations.
This chapter demonstrates how to (a) assess the direction and size of a correlation, (b) identify limitations of correlation, and (c) calculate the most common form of correlation: r, the Pearson correlation coefficient. We then use the six steps of hypothesis testing to determine whether a correlation is statistically significant.