Over at The Hardest Science, Srivastava is gearing up to teach a course on SEM and thinking about how to warn students against the pixie dust approach to modeling and analysis.
This is definitely not easy, and something I struggle with, even when I have the time to prepare and know better. Last week I was helping someone learn how to do multi-level modeling in SPSS, a software package that apparently I am supposed to be competent with. I was so happy when I figured out the MIXED function to get the output I wanted, that this is all I was able to focus on. I totally forgot about conveying the higher level concepts of model formulation, such as writing out equations for each of your models and plotting regression lines for each of the subjects or treatment levels to get a feel for where modeling energy should be expended. Part of the problem is that I don’t know anything about plotting in SPSS. A poor excuse, I know.
Our tools do constrain us, and while knowing the principles and concepts is important, there is another aspect of statistical literacy, which is making your tools do what you want.
Photo cc-by FordRanger
-
differential posted this
