CODEWITHSUNDEEP
rsasloess
Elderstahl
Elderstahl

Reputation: 33

Bootstrapped data loess smoothing in SAS and R

After using loess.smooth function in R on bootstrapped data, the bootstrapped dataset (n = 134560) reduced to (n = 50) observations. is there any way to replicate this effect in SAS?

I am not very much familier with PROC LOESS. I tried a few attempts but did not understood how can I replicate this effect (effect = final dataset of reduced n).

The reason for this question is: I find bootstrapping in SAS much easier, but not able to make a plot-able dataset in SAS.

Any help or direction in this regard is much appreciated. Thanks.

Upvotes: 1

Views: 287

Answers (1)

Joe
Joe

Reputation: 63434

There may well be a better solution, but at minimum:

First, run PROC LOESS and grab the output dataset.

proc loess data=sashelp.enso;
   model pressure = year /details(outputstatistics);
   output out=loess;
run;

Now, run PROC SURVEYSELECT to pull at random from that, or even just take every Nth row fro that result dataset. I imagine r's loess function does something similar, although it is probably more intelligent than that (choosing local minima/maxima of the loess curve for example).

Upvotes: 0

Related Questions