Sydney GordonShow up the bird place by answering the question they didn’t?
Requesting #stata help: can anyone tell me why mkspline supports linear and cubic, but not quadratic, estimation? Is this a quirk or some theory I’m missing?
Cameron Pfiffer@sydneygordon Probably because quadratics are even functions and have some weird behavior when doing curve fitting
@sydneygordon Shoot, forgot to link this http://www.biology.arizona.edu/biomath/tutorials/polynomial/Polynomialbasics.html
Stas Kolenikov@cameron_pfiffer @sydneygordon the cubics do have some theory -- it is the optimal functional form when you attempt to minimize the overall curvature (the integral of the squared second derivative). Something like this may help: https://math.stackexchange.com/a/699218/97144
