Friday, August 9, 2013

how would you frame...

...really sensible, but small, coefficients and awful pseudo R2s?

I'm not going to try to shy away from anything here. The results match very well to theory, they just don't matter all that much. My discussion says that theory seems to explain behavior on the margin well but there area lot of non-economic contributors to this outcome that make it's ultimate magnitude insignificant.

It's more significant for some of the models, but the ones I care about less.

So I want to be honest but I just want to make sure I'm not still being too generous to my own biases.

Statistical significance is not a problem in any of these cases.


  1. Is it possible that the model sort of works at the margin because at the margin the underlying reality can be effectively modeled by a linear approximation? Away from the margin non-linear effects dominate? Just a thought.

  2. Perhaps frame it as statistically significant but not practically relevant.


All anonymous comments will be deleted. Consistent pseudonyms are fine.