Closed andreasnoack closed 1 year ago
I'd say that the purpose of LOESS is to provide smooth predictions but that is not the case with the current implementation. Compare the LOESS lines produced by this package to the one from R. The tuning parameters should be identical.
julia> using Loess, CairoMakie julia> lf = loess(df.x, df.y); julia> f, ax, l1 = scatter(df.x, df.y); julia> lines!(ax, px, Loess.predict(lf, px), color=:black); julia> f
> df <- read.csv("/Users/andreasnoack/Downloads/loessdf.csv") > plot(y ~ x, df) > lf <- loess(y ~ x, df) > lines(px, predict(lf, px))
loessdf.csv
closed by https://github.com/JuliaStats/Loess.jl/pull/63 ?
Yes!
I'd say that the purpose of LOESS is to provide smooth predictions but that is not the case with the current implementation. Compare the LOESS lines produced by this package to the one from R. The tuning parameters should be identical.
Loess.jl:
R:
Data
loessdf.csv