On Correlations and Causation

from http://xkcd.com/552/

As you heard in our second lecture (and, undoubtedly, have heard before in previous courses in Psychology and other scientific fields), just because a correlational relationship emerges between variables in a study does not mean we know anything regarding the causal relationship between those variables.

Check out this popular media write-up of a research study.  The title of the web piece is Bullied Kids Averse to Exercise and it describes a relationship between bullying and decreased physical activity (i.e., a negative correlation).

The topic here is interesting to begin with, as bullying certainly has social psychological antecedents and implications.  But let’s just focus on research methods for the time being.  Any problems with the way the research findings are described in the story on outsideonline.com?  That is, does the web piece offer conclusions that are appropriate based on the type of study being addressed?  What do you think?

This entry was posted in Chapter 2, Media, Research Methods, Research Study. Bookmark the permalink.

9 Responses to On Correlations and Causation

  1. I wonder why this study only polled students from 6 mid western elementary schools? Any only 6 schools for that matter? It seems like there would be a bias based on the standards of schools from one region and a small sample size. For example, maybe these schools do not put a strong emphasis on physical education, therefore skewing the results from the start.

    • Tory Kolbjornsen says:

      Agreed, since these tests were so localized it seems like there could be links to kids in the Midwest not valuing athleticism, or that gym teachers in that area are all awful. Additionally it seems like trends in 4th and 5th graders would likely be the same since they are still in the same school with the same peers and likely the same teachers, if their environment hasn’t changed at all why would their attitude?

      • Olympe Nalbandian says:

        I agree, the environment of 4th and 5th graders in schools with the same general geographic area is quite contained and doesn’t include many differences overall. The researcher should have widened the geographical areas for the students, as well as the age range. I think that if the researcher studied a more diverse set of students, the results (positive or negative correlation) would have been more compelling.

  2. Olympe Nalbandian says:

    I do not find the correlation found to be all that remarkable. To begin with, the study only tested a small number of children from only six schools and typically, a greater number of test subjects yields more data that can better show overall trends. Also, bullying is a major problem that effects almost all people at some point during childhood and can cause someone a variety of physical and psychological problems aside from aversion to exercise. In the same way, a lack of exercise among youth is a major cause of the obesity epidemic and has many possible causes by itself. I just don’t think this study included enough variables and manipulations to really connect the two in a convincing way.

    • I agree, it seems that there are many other factors that could have contributed to the results described by this study. Similar to what you were saying, bullying in general could cause the psychological effects that lead to aversion from exercise, whether or not the bullying occurred during P.E. class. In addition, based on my understanding of bullying, especially at such a young age, it is highly probably that the students being bullied during P.E. were also bullied in other situations. I would be interested to hear how the study was able to distinguish the effects of P.E. bullying versus bullying in other contexts.

    • ahoi01 says:

      I agree; it seems premature to make this claim when the sample size is so small and regional. It would’ve been interesting to know the methods behind participant selection, because there is potential for selection bias. However, I think this study may serve as an interesting spring-board for other researchers to further investigate — particularly, the effect of bullying on exercise habits of average weight children.

  3. Ki Jung Lee says:

    I do not think measuring weight was a proper measurement for obesity or being over weight without considering individual’s height. Also, the study does not define what it means to be obese or over-weight. Is it based on the BMI or ideal body weight formula or other? Study should define bullying as well. It is possible that the article only reported summary of the study and did not provide details of the studies like operational definition of each variables in the study. Another problem that I noticed in the study was information on the sex. Were the participants randomly selected on different sexes? I believe sex would be another variable that should be controlled to appropriately reflect the real world. I believe sex could become a confounding variable that makes the study’s conclusion invalid. Other peers already mentioned about the small pool number that does not really provide statistical importance on the data. All of these reasons led me to believe that study’s conclusion was inaccurate and its correlation is biased.

    • I agree and not only did the study not define exactly what bullying was it did not specify what the bullying was centered around. Were the children bullied mostly about their physical appearance or any other trait and why would that be reflected in their aversion to sports. As others have also mentioned the size of the study is rather small while looking at a small pool of the population in both age and geography. The site also cites the solution as simply anti bullying strategies to raise youth physical activity in general while helpful I do not believe it is the only factor that would increase it.

    • I was wondering how bullying would be defined. It would be very hard to specify especially if the type of bullying would go beyond physical bullying. It would be hard to scale severity of bullying but also very important as well. It would also be important to look at frequency of bullying. Daily bullying with a specific person would look very different to more dispersed bullying with varying perpetrators. I would also want to look at the nature of the bullying. If it were based on socially significant identifiers this would be significant. I agree that the correlation is heavily biased and this correlation would have to be reevaluated to be taken seriously.

Leave a Reply

Your email address will not be published. Required fields are marked *