A post at the message board brought my attention to a method of statistical analysis that relies on a broadly Aristotelian framework rather than null hypothesis testing. I must say that this has me quite excited. The standard null hypothesis testing uses extremely complex statistical methods that seem to abstract away from reality with every step. For instance, they rely on aggregate averages of a sample compared to estimated population averages. The issue for me is whether an average of a group is even a meaningful thing to begin with. It seems to me that when you’re comparing the average of a treatment group to the average of a control group, you’re comparing two things that don’t actually exist in reality. The people exist, the observations of those people exist, but I’ve never seen a good argument as to why it is meaningful to create an average of those people, or compare that to an “estimated” population average. There might be a good argument for why that’s meaningful, but I haven’t seen it, and I’m guessing that psychology as a whole hasn’t actually stepped back and considered that question for a half century or more. There are of course many more arguments that make a case against null hypothesis significance testing, but never have I seen a better alternative.
Which is why Observation Oriented Modeling seems quite fascinating to me. Here is a book describing it and here is the website for their research team. Grice is not only an Aristotelian, he is also interested in psychology of religion. I really ought to apply to the Oklahoma State program! It is not a refinement of null hypothesis significance testing, it is a new method that operates on different, Aristotelian, assumptions.
The only problem is that it is so radically different than the statistical methods I’ve just been trained with, that it’s incredibly difficult for me to wrap my head around. I need a “For Dummies” version. The other issue I have is that some more vetting and groundwork will have to take place before I would ever be able to publish using OOM. For that reason, I am hoping this conversation takes place soon so I can really consider using OOM in research. I would love to, especially if it places the emphasis in research on building accurate models instead of comparing average statistical scores. When I scrounge up some money I’ll get the book