From "Grand Theft Auto" to Inferential Statistics

A Research Methods student from last semester, Jordan Prendez, sent me the following link: http://www.gamespot.com/news/6147420.html. It is a short article describing a research study showing that players of a more violent video game are more prone to violence themselves than players of a less violent video game.

First, let me point out that the study, as described, is a good example of a two-group randomized experiment with multiple dependent variables. It is an experiment because it has a manipulated independent variable and the researchers appear to have tried to control other variables.

Second--and this was Jordan's point--it is interesting how the gamers who commented on the article attacked the research (presumably without reading the original study). This is reminiscent of a famous study by Lord, Lepper, and Ross (1979), showing that people are good at finding the flaws in a research study ... but only when they already disagree with the study's conclusions.

Statistics students, in particular, should note the argument that appears in a couple of the comments. In essence, "There's no way that a sample of 100 [the number of participants in the study] can represent the whole population of gamers." Intuitively, this seems right because 100 is a very small fraction of the total number of gamers out there. If there were 10 million gamers in the population, for example, this sample would only represent 1/100,000th of that population. But this intuition is wrong. A sample of 100 will generally be much more similar to the population--even a very large population--than most people realize. And this is what the field of inferential statistics is all about.

Thanks, Jordan!