# Is it correct to say the "smooth curve fitted by LOESS"?

I use the package scatter.smooth in R, and I want to have a figure text describing the line in the scatterplot for my report.

Is is okay to say "the line is a Smooth Curve Fitted by LOESS (a reference to the package)" or need I explain it more? (It's a software engineering master thesis)

Otherwise, how can I describe what the line is in one short sentence?

• Including (as you plan to do) a precise reference to the package or, possibly, to a publication describing the technique is a very good idea. Even if readers don't understand a part of your description or don't know the method, they have an unambiguous way to find out what you did. Other people might have ideas on how improve the description but even a very good description is less helpful than a good reference IMO.
– Gala
Apr 26 '13 at 11:44
• Because "smooth" in mathematics (as an adjective) means something that the loess curve is not, it might be clearer for some readers to use "smooth" as a noun, as in "the curve is a loess smooth."
– whuber
Apr 26 '13 at 14:37

If you just say "loess" people probably won't know what you mean. Perhaps "locally weighted regression" is better? The R help description of loess in stats is

Fit a polynomial surface determined by one or more numerical predictors, using local fitting.

• localy weighted regression sounds good. The description in the help is too technical for the figure text. But I'll keep the reference if someone is not satisfied with the simpler text. thx Apr 26 '13 at 13:30

In a master's thesis, you should probably explain a bit what the method does. In addition, you should somewhere give the details (package, version, function name, parameters). This may be all in the text, or shortly in the text and full details in an appendix.

In papers, where text needs to be shorter, I e.g. say

Data analysis was performed in R [1] ... smoothing interpolation by function loess ...

• "smoothing interpolation" is closer to how I use loess mainly - for you locally weighted regession may be the better explanation.
• stats is one of the core R packages, so I don't spell it out. Other packages I give with citation.