Tuesday, 22 January 2013

Natural Experiments and Quasi-Experimental Design

When reading chapter four in Luker, her use of the term “natural experiment” (pg. 57) grabbed my attention. I had heard the term before when reading Jared Diamonds Gun, Germs and Steel. Thinking back I remembered Diamond arguing that as there was no possibility to create a traditional experiment to determine causal history of societies in the distant past, so he looked to “natural experiments” to determine what variables could have affected the societies. This made me think when reading Luker’s views on natural experiments and the quasi-experimental method used by social scientists. A social scientist cannot control all of the variables in the way that the traditional use of the scientific method calls for. So social scientists use their quasi-scientific method to try to account for all the variables that could affect their result. As Luker states, variables are not always accounted for properly. She demonstrated this though describing the issues in a Harvard Nurses’ Health Study on Hormone replacement therapy. The results of this study were skewed as they did not account for the variable of the nurses’ health, as opposed the health of the greater community. To deal with this issue of unaccounted-for variables, and by extension, flawed results, Luker seems to suggest that a researcher should be flexible and choose to keep parts of experimental methodologies that are most relevant to the project. In this way the researcher will not get caught up in methodology, but be able to focus on the variables and how they may affect the outcomes (Luker pg. 57-61). This may alleviate the issue, but it is impossible to account for every variable, or to know without a doubt what variables are relevant to your question. So the chapter left me wondering how I will best account for all relevant variables and not be the one to miss the gorilla in the room.

1 comment:

  1. Ah, yes. And the NOISE in the Shannon-Weaver model of communication rears its chaotic head! I suppose the hope is that once you've narrowed down your research topic, the method most appropriate to that topic, that is, the method most likely to account for all of the relevant variables, will become apparent....that's the hope, at least!