feature request: loess/lowess regression

[koaning]

I've always wanted to have loess regression in python. R has a cool version of it but it has always been missing. This would be a great model to host here.

[koaning]

The easiest way to explain what it needs to do:

[dvanamstel]

DIBS

[tsuhina]

Second that

[koaning]

library(tidyverse)

df <- data_frame(xs=1:120, ys=10*sin(xs/10) + rnorm(120, 0, xs/10)) %>% 
  filter(ifelse((xs < 100) & (xs > 75), FALSE, TRUE))

ggplot() + 
  geom_point(data=df, aes(xs, ys)) + 
  geom_smooth(data=df, aes(xs, ys), span=0.5)

[koaning]

ggplot() + 
  geom_point(data=df, aes(xs, ys)) + 
  geom_smooth(data=df, aes(xs, ys), span=0.1)

[koaning]

ggplot() + 
  geom_point(data=df, aes(xs, ys)) + 
  geom_smooth(data=df, aes(xs, ys), span=1)

[koaning]

http://web.ipac.caltech.edu/staff/fmasci/home/astro_refs/LocalRegressionBook_1999.pdf

[dvanamstel]

Nice book, thanks!

I expect to have a few hours next week to finish up what we started.

[koaning]

It seems to be in statsmodels. Maybe it is not a right fit.