craggit vs. churdle

[Elina Strawberry]

Hi, could you please help me to understand the difference between the craggit command and the chrudle command? From what I understand with churdle you can choose either linear or exponential estimation, which you cannot in the craggit. But is that the only difference? Is one command more recent (better)? If no difference I would choose the churdle estimator because of the convenient post estimation... But not comfortable to leave craggit behind before I have understood the difference between the two.Are the estimators used for the exact kind of estimation and data?

Thank you in advance!
Elina

[Jeff Wooldridge]

I think churdle is more recent and more flexible. It allows factor notation.

As an aside, Stata should be a little more careful with its label of "linear" because the conditional mean is not linear. It has the truncated normal form, which is a linear term plus a function of the inverse Mills ratio. By contrast, conditional on y > 0, the mean is really exponential in the exponential case. The reason I mention this is that the linear function is never used for prediction or for computing marginal effects; it is the truncated normal form (with the "linear" option) or the exponential function.

JW