If one is trying to improve education given class sizes of 1 or 2 dozen students, it may not be helpful to pursue better quantitative studies for class sizes of 1 or 2 students.
The goal, in my view, of such research is to make discoveries. I'm dubious of academic research that targets scalable educational methods too closely.
In essence, I agree with Patrick. I want to know if "2-sigma" is true, replicable. Establish that watermark. At that point we'll have a high watermark for education, what a median student can achieve in a near ideal environment. Is there really a 2 sigma difference between normal school results and ideal?
If so, this difference represents potential.
Figuring out how to apply that IRL with resource constraints, scaling issues, etc... That may not be a job for academics. Either way, it comes later.
