How does the "smoothing factor" in smoothdata work?

Dear Matlab community,

I am currently trying to remove noise in sensor data with the smoothdata function. I noticed that I usually get far better results by choosing a "smoothing factor" instead of a moving window of values for median calculation (or other methods). However, I need to provide mathematically rigid results, so I wonder how this smoothing factor works. All that I found in the help document was that it "adjusts the level of smoothing by scaling the heuristic window size".

Does anyone of you know what the "heuristic window size" is, if it changes throughout the data series and if I can determine its size?

Many thanks in advance,

Ed

[Brian Hegger]

Did you ever figure this out? I am noticing using a very small "smoothing factor" gets a fairly good approximation.

[Scott MacKenzie]

Just to clarify, are you interested in knowning what the heuristic is when using the 'smoothingfactor' option or in knowing what actual window size is used?

[DGM]

FWIW, you can always look at smoothdata.m and its support files. On line 436 (in R2019b), smoothdata.m sets winsz by calling $MLROOT/toolbox/matlab/datafun/+matlab/+internal/+math/chooseWindowSize.m

The file can be plainly read and has plenty of comments if you want to figure it out.