Everyday Research Methods

Studying video games in the lab and in the real world.

A story in this month's New York Times summarizes a bunch of studies that address the question of whether video games cause violent behavior. Read the story here.

The article summarizes research from psychology, sociology, and criminal justice scholars. Some of the studies are correlational, and others are experimental.

Here's one study described in the article :

In one recent study, Christopher Barlett, a psychologist at Iowa State University, led a research team that had 47 undergraduates play “Mortal Kombat: Deadly Alliance” for 15 minutes. Afterward, the team took various measures of arousal, both physical and psychological. It also tested whether the students would behave more aggressively, by having them dole out hot sauce to a fellow student who, they were told, did not like spicy food but had to swallow the sauce.       

Sure enough, compared with a group who had played a nonviolent video game, those who had been engaged in “Mortal Kombat” were more aggressive across the board. They gave their fellow students significantly bigger portions of the hot sauce.

a) What kind of study is this--an experiment, a quasi-experiment, or a correlational study? Give the most precise answer you can.

Here is another study described in the article: 

Dr. Ward and two colleagues examined week-by-week sales data for violent video games, across a wide range of communities. Violence rates are seasonal, generally higher in summer than in winter; so are video game sales, which peak during the holidays. The researchers controlled for those trends and analyzed crime rates in the month or so after surges in sales, in communities with a high concentrations of young people, like college towns.       

“We found that higher rates of violent video game sales related to a decrease in crimes, and especially violent crimes,” said Dr. Ward, whose co-authors were A. Scott Cunningham of Baylor University and Benjamin Engelstätter of the Center for European Economic Research in Mannheim, Germany.

b) What kind of study is this--an experiment, a quasi-experiment, or a correlational study? Give the most precise answer you can.

Finally, here's one more statement:

Some studies in schools have found that over time digital warriors get into increasing numbers of scrapes with peers — fights in the schoolyard, for example. In a report published last summer, psychologists at Brock University in Ontario found that longer periods of violent video game playing among high school students predicted a slightly higher number of such incidents over time.

c) What kind of study is this--an experiment, a quasi-experiment, or a correlational study? Give the most precise answer you can.

Interestingly, the journalists includes this statement in his piece:

[none of the scientists] interviewed in this article, receive money from the gaming industry. 

d) Why might funding of the research be important?

e) How might we reconcile the results of these three studies, which seem to draw opposite conclusions? 

A blog about interactions

You don't see people blogging about interaction effects, at least not every day. I just ran into this 2011 Ben Goldacre column about interactions, which he describes in my favorite way, as a "difference in differences." Take a look here. Instructors, this might be a handy example for teaching Chapter 11.

 

No effect of day care on child behavior...after all?

Does time in day care put children at risk for behavioral problems? The latest study, released recently in the empirical journal Child Development, suggests that the answer is, "probably not."

The online source Science News summarizes some of the key points of this study:

A new study that looked at more than 75,000 children in day care in Norway found little evidence that the amount of time a child spends in child care leads to an increase in behavioral problems, according to researchers from the United States and Norway.

a) The following description suggests that past researchers have used multiple regression in their studies of day care and behavior. Read the following paragraph carefully, jotting down what earlier U.S. studies used as dependent and predictor variables in their regression.

...the researchers took issue with the common approach, which is to compare children from different families who spend varying amounts of time in child care because of family choices. Although earlier U.S. studies using this method tried to control for parent and family characteristics -- such as income and education, mental health and intelligence -- the method leaves open the possibility that differences between families in areas other than child care choices are, in fact, the true causes of behavior problems.

The last line of that paragraph, "the method leaves open the possibility that differences between families in areas other than child care choices are, in fact, the true causes of behavior problems" is a good practical example of the message in Chapter 8 (p. 225).  Regression cannot establish causation, even when it controls for many possible outside variables. What are some of the outside variables that might vary across families, besides income, education, and intelligence?

b) Below is a description of how this new study did things differently. Read the description, and then, in your own words, explain why it might it be important to study kids within the same family.

Given the scope the Norwegian data, the researchers were able to compare children who came from the same families but who spent varying hours in child care, effectively resolving the issue of external influences. When they did this, they found no statistical evidence to point to increased behavioral problems. Siblings who spent more time in day care exhibited the same behavior as siblings who spent less time in day care, Dearing said.

c) Finally, you can also apply the second half of Chapter 10, which discusses null effects. When a study finds no effect of one variable on another, there are a number of possible explanations. Can we be sure that the null effect in the day care study is not simply caused by large within-groups variability?  Part of the answer has to do with the sample size. Large samples can help cancel out unsystematic variability between groups. 

Large samples also have a lot of statistical power:

[Dearing, one of the researchers said,]  "With such a very large sample, even very, very small correlations would be statistically significant. But we found no association in our most sophisticated models."

Very large samples are like huge, powerful flashlights. They are able to detect small effects, if they exist (see p. A33). So if a large sample doesn't find an effect, then most likely, it's just not there.