To this date I must admit that I have struggled with the intuition behind models and the statistical analysis that backs up the research that tests whether or not a model is valid. The earliest I can recall this struggle was when I read about chaos theory when it presented in the book Jurassic Park by Michael Crichton. Was it self-evident that Jurassic Park would fail because a butterfly flapped its wings on the other side of the world? The fascinating idea of chaos theory and emergent behavior of models even led me to my senior research paper in college about the non-linear distribution of bird morphology. Unfortunately even then my mathematical and statistical skills were not up the standard I needed to present an effective paper. I just couldn’t dig deep enough into the material. It’s also likely why my pursuit for continuing research at the graduate level stalled.
What makes it worse is that despite my limitations with statistics and probability theory, I will go out on a limb and state that when it comes to research (at least in certain domains) what is actually being tested is often not what is being purported as the intuition behind the tested model or what is being tested is very limited and thus of little use. Genius is on display when someone can line the two up and testing is very clear and carries widespread implications. However genius is rare. While I strive for a goal clearly less than genius status, I do want to master the tools being used to evaluate the quality of someone’s research so that I can recognize true genius.
Before I ramble on too long, the intent of this blog post was to bring up probabilistic graphical models. It is my belief that graphical models which became popular around the time Jurassic Park was published is an alternative path to doing research that might provide me with a new way of approaching models. It might help clear up long-standing arguments in evolutionary biology. More importantly, graphical models get closer to understanding what generates the signal when looking at noisy data. It gets at the question why? Developing a graphical model might be a more effective tool to understand why Jurassic Park ended up failing (and why investors should never have invested) or why birds evolved into certain sizes but not others.
Onwards to mastery.
