Ken Blake, Ph.D.

Comparing a variable's percentages with percentages you've specified: The one-sample chi-square test.

Dixon and Linz (2000) set out to learn, among other things, how well the prevalence of white victims in Los Angeles television news coverage of homicides matched up with the prevalence of white victims in actual homicide data. They suspected that white victims would be overrepresented in coverage - evidence of a bias that could lead to all sorts of public misperceptions about homicide. But how could such bias be tested for in an objective, transparent and replicable way?

According to actual crime data that Dixon and Linz obtained for the Los Angeles area at the time of their study, about 20 percent of homicide victims were black, about 39 percent were Latino, about 4 percent were some other race, and about 9 percent were white. So, Dixon and Linz collected a representative sample of 139 homicides reported in Los Angeles-area television news broadcasts and reasoned that, if the coverage were unbiased, about .20 x 139 = 28 of the victims would be black, about .39 x 139 = 54 would be Latino, about .04 x 139 = 5 would be some other race, and about .09 x 139 = 13 would be white. The found, as you might guess, that quite a bit of difference existed between these numbers and what the local television news coverage portrayed.

This video describes data from Dixon and Linz's study and shows how to analyze it using an Excel pivot table and a one-sample chi-square test.

The test can tell you whether the difference between the percentages observed in TV news content and the percentages found actual crime data are large enough to suggest that the difference occurred for some nonrandom reason.

Credit for the idea of using Dixon and Linz's (2000) data to illustrate this statistical technique belongs to Hayes (2005).

Want to practice what you saw in the videos? Here's the dataset. Note: Be sure to save the Excel file on your own computer before trying to follow along with the video. Depending on your computer's setup, you might get an error if you just click the link, let your computer open the file automatically, then try to start working without first saving the file.

References:

Dixon, T.L., & Linz, D. (2000). Race and the mispresentation of victimization on local television news. Communication Research, 27, 547-573.

Hayes, A.F. (2005). Statistical methods for communication science. Mahwah, NJ: Lawrence Erlbaum.