Hello, I have a question concerning the weights argument of the loess function in the stats package. Do the weights correspond to multiplying the local regression equation by the weights or by the square root of the weights (like in lm). In "http://www.netlib.org/a/cloess.pdf" page 7 it appears that the former is true. A look at the source code loess.f lets me conclude that the latter is true. Since I do not know Fortran, I would be really delighted if someone could clarify this. Best, Felix