Thursday, September 23, 2010

Precision, Generality and Realism

It is nearing the end of the second week of classes and I am in the process of designing a project for my M.Sc. thesis. A couple of days ago I met with my supervisor to discuss some potential ideas for my project. It is not uncommon for him to rant about all sorts of different ideas; however, this time something he said really sparked a fire in me. He pointed out that my B.Sc.H thesis was a natural experiment that involved no set-up or manipulations of any kind and that he really appreciates the beauty of these “realist” experiments- an experiment where you simply go out and record or collect exactly what you see. This inspired me to do a literature review of this idea of natural experiments and I came across a very interesting paper in which Richard Levins addresses different types of biological studies.

In 1966, Levins wrote a paper titled Strategy of Model Building in Population Biology published in American Scientist. He explained that it is ideal to work with mathematical models in biology that maximize three things: 1) Generality, 2) Realism and 3) Precision, but that it is never possible to maximize these three things simultaneously.

Levins suggests that there is a trade-off between these three ideals and that three strategies have evolved to deal with this problem. For example, one could sacrifice generality to realism and precision. The focus of an experiment could be on the behaviour of an organism within very small parameters that would yield real and precise results but would tell us very little about the organism’s behaviour in general. Scientists could also sacrifice realism to generality and precision in which general equations and models can be developed that yield precise results but are extremely unrealistic. Finally one can sacrifice precision to realism and generality which rely on flexible models that assume wide variance in different functions.

Different scientists prefer different methods and each way has its merit. In my blog, I will discuss current studies in ecology that sacrifice generality or precision for realism and why I believe this approach is the most valuable.


Source: Levins, Richard. 1966. The Strategy of Model Building in Population Biology. American Scientist 54(4): 421-431.