credible_gradient: Function for plotting posterior regression line

In stablemarkets/ChiRP: Chinese Restaurant Process Mixtures of Regressions

Description

This function takes as input some continuous covariate x and predictions from NDPMix, ZDPMix, fDPMix, etc. It outputs a credible band and posterior mean line of the predictions across x

Usage

credible_gradient(
  x,
  post_draws,
  col_gradient = "lightblue",
  col_mean_line = "steelblue"
)

Arguments

x	A vector of type as.numeric to plot on the x dimension of the plot.
post_draws	matrix with dimensions length(x) by # posterior draws.
col_gradient	Optional. Defaults to lightblue. Darkest color that gradient goes to.
col_mean_line	Optional. Defaults to steelblue. Color of the posterior mean regression line.

Details

Please see https://stablemarkets.github.io/ChiRPsite/index.html for examples and details.

Value

overlays credible band and posterior mean line.

Examples

set.seed(1)

N = 200
x<-seq(1,10*pi, length.out = N) # confounder
y<-rnorm(n = length(x), sin(.5*x), .07*x )
d <- data.frame(x=x, y=y)
d$x <- as.numeric(scale(d$x))
d$y <- as.numeric(scale(d$y))

d_test = data.frame(x=seq(max(d$x), max(d$x+1 ), .01 ))


res = fDPMix(d_train = d, d_test = d_test, formula = y ~ x,
             iter=100, burnin=50, tau_x = c(.01, .001) )



par(mfrow=c(1,1))
plot(d$x,d$y, pch=20, xlim=c(min(d$x), max(d$x)+1 ), col='gray')

credible_gradient(d$x, res$train)
credible_gradient(d_test$x, res$test, col_gradient = 'pink', col_mean_line = 'darkred')

points(d$x, d$y, pch=20, col='gray') ## re-plot points to bring the to the top.
abline(v=max(d$x), lty=2)

stablemarkets/ChiRP documentation built on July 26, 2021, 2:25 a.m.