Residuals from lowess curve I am trying to obtain the residuals from a lowess fit. I’m using the lowess( ) function. Is there a way to do this? 
 A: Not directly, but you can get the information, usually quite readily.
The closely related loess function in R does work with predict (see ?loess), you might find that easier.
Typically, lowess evaluates the function at the data (so you can just sort the output and find the $y$ for your specific $x$), but for some settings of the input arguments it may skip some points, and you may want to avoid the possibility of duplicated $x$ values.
One way to do this fairly cleanly is to use approxfun to define a function, and then you don't have to worry what the settings were or whether there are omitted or duplicate $x$'s:
 # first set up a lowess fit:
 lfit <- lowess(x,y)

 # create a functional version of the lowess fit
 lfun <- approxfun(lfit)
 fitted <- lfun(x)
 resid <- y-fitted

Let's try an example:
 carsfit <- with(cars,lowess(speed,dist))
 carslowessfun <- approxfun(carsfit)
 fitted <- lfun(cars$speed)                          #$
 resid <- cars$dist-fitted                           #$
 plot(fitted,resid)
 abline(h=0,col=8)

$\ $
