I recently ran across the blog, correlated.org, as well as a book that goes with it, called Correlated (both are by Shaun Gallagher). Every day, the author posts the results of a correlation that is based on survey data. The goal of the project is to entertain. Many of the correlations are amusing, such as the correlation between liking Disney movies and preferring lemonade to iced tea (see figure).
The correlations on correlated.org are presented very differently from the ones in the textbook. The text shows you two ways to present correlational data: a scatterplot and a bar graph. You use scatterplots when both variables are quantitative, and you use a bar graph when one is quantitative and the other is categorical.
In contrast, correlated.org uses two pie charts to represent each of its correlations. But for some reason, pie charts are almost never used in psychology research.
Think about the two variables in the pictured correlation. The first variable (is your favorite animated movie a Disney movie?) is categorical--the answers are yes or no. Similarly, the second variable (do you prefer iced tea or lemonade?) is also categorical--the answers are iced tea or lemonade. When both variables are categorical, a scatterplot can't be used. Pie charts are one option, but bar graphs are possible here too (and are more elegant, in a way, since you can use one bar graph in place of two pie charts).
a) Why does this count as a correlational study and not an experiment or a survey?
b) You can present the lemonade/Disney results as a single bar graph rather than a pie chart. Sketch it out for yourself.
c) Think about the causal direction of the lemonade/Disney results. Is it clear which of the two variables comes first in time? What third variables seem likely to be associated with both variables?
d) What are the advantages and disadvantages of using two pie charts versus one bar graph? Which one is easier to understand? Which one serves author Gallagher's purpose?
e) What questions would you ask to determine the external validity of this result?