R: Visually Weighted Regression.

vwReg {petersenlab}

R Documentation

Visually Weighted Regression.

Description

Create watercolor plot to visualize weighted regression.

Usage

vwReg(
  formula,
  data,
  title = "",
  B = 1000,
  shade = TRUE,
  shade.alpha = 0.1,
  spag = FALSE,
  spag.color = "darkblue",
  mweight = TRUE,
  show.lm = FALSE,
  show.median = TRUE,
  median.col = "white",
  shape = 21,
  show.CI = FALSE,
  method = loess,
  bw = FALSE,
  slices = 200,
  palette = colorRampPalette(c("#FFEDA0", "#DD0000"), bias = 2)(20),
  ylim = NULL,
  quantize = "continuous",
  add = FALSE,
  ...
)

Arguments

`formula`	regression model.
`data`	dataset.
`title`	plot title.
`B`	number of bootstrapped smoothers.
`shade`	whether to plot the shaded confidence region.
`shade.alpha`	whether to fade out the confidence interval shading at the edges (by reducing alpha; 0 = no alpha decrease, 0.1 = medium alpha decrease, 0.5 = strong alpha decrease).
`spag`	whether to plot spaghetti lines.
`spag.color`	the fitting function for the spaghettis; default: `loess`.
`mweight`	logical indicating whether to make the median smoother visually weighted.
`show.lm`	logical indicating whether to plot the linear regression line.
`show.median`	logical indicating whether to plot the median smoother.
`median.col`	color of the median smoother.
`shape`	shape of points.
`show.CI`	logical indicating whether to plot the 95% confidence interval limits.
`method`	color of spaghetti lines.
`bw`	logical indicating whether to use a b&w palette; default: `TRUE`.
`slices`	number of slices in x and y direction for the shaded region. Higher numbers make a smoother plot, but takes longer to draw. I would not set `slices` to more than 500.
`palette`	provide a custom color palette for the watercolors.
`ylim`	restrict range of the watercoloring.
`quantize`	either `continuous`, or `SD`. In the latter case, we get three color regions for 1, 2, and 3 SD (an idea of John Mashey).
`add`	if `add == FALSE`, a new ggplot is returned. If `add == TRUE`, only the elements are returned, which can be added to an existing ggplot (with the `+` operator).
`...`	further parameters passed to the fitting function, in the case of loess, for example, `span = .9`, or `family = "symmetric"`.

Details

Creates a watercolor plot to visualize weighted regression.

Value

plot

Examples


# Prepare Data
data("mtcars")
df <- data.frame(x = mtcars$hp, y = mtcars$mpg)

## Visually Weighted Regression

# Default
vwReg(y ~ x, df)

# Shade
vwReg(y ~ x, df, shade = TRUE, show.lm = TRUE, show.CI = TRUE,
quantize = "continuous")
vwReg(y ~ x, df, shade = TRUE, show.lm = TRUE, show.CI = TRUE,
quantize = "SD")

# Spaghetti
vwReg(y ~ x, df, shade = FALSE, spag = TRUE, show.lm = TRUE, show.CI = TRUE)
vwReg(y ~ x, df, shade = FALSE, spag = TRUE)

# Black/white
vwReg(y ~ x, df, shade = TRUE, spag = FALSE, show.lm = TRUE, show.CI = TRUE,
bw = TRUE, quantize = "continuous")
vwReg(y ~ x, df, shade = TRUE, spag = FALSE, show.lm = TRUE, show.CI = TRUE,
bw = TRUE, quantize = "SD")
vwReg(y ~ x, df, shade = FALSE, spag = TRUE, show.lm = TRUE, show.CI = TRUE,
bw = TRUE, quantize = "SD")

# Change the bootstrap smoothing
vwReg(y ~ x, df, family = "symmetric") # use an M-estimator for
# bootstrap smoothers. Usually yields wider confidence intervals
vwReg(y ~ x, df, span = 1.7) # increase the span of the smoothers
vwReg(y ~ x, df, span = 0.5) # decrease the span of the smoothers

# Change the color scheme
vwReg(y ~ x, df, palette = viridisLite::viridis(4)) # viridis
vwReg(y ~ x, df, palette = viridisLite::magma(4)) # magma
vwReg(y ~ x, df, palette = RColorBrewer::brewer.pal(9, "YlGnBu")) # change the
# color scheme, using a predefined ColorBrewer palette. You can see all
# available palettes by using this command:
# `library(RColorBrewer); display.brewer.all()`
vwReg(y ~ x, df, palette = grDevices::colorRampPalette(c("white","yellow",
"green","red"))(20)) # use a custom-made palette
vwReg(y ~ x, df, palette = grDevices::colorRampPalette(c("white","yellow",
"green","red"), bias = 3)(20)) # use a custom-made palette, with the
# parameter bias you can shift the color ramp to the “higher” colors
vwReg(y ~ x, df, bw = TRUE) # black and white version
vwReg(y ~ x, df, shade.alpha = 0, palette = grDevices::colorRampPalette(
c("black","grey30","white"), bias = 4)(20)) # Milky-Way Plot
vwReg(y ~ x, df, shade.alpha = 0, slices = 400, palette =
grDevices::colorRampPalette(c("black","green","yellow","red"),
bias = 5)(20), family = "symmetric") # Northern Light Plot/ fMRI plot
vwReg(y ~ x, df, quantize = "SD") # 1-2-3-SD plot

[Package petersenlab version 1.0.0 Index]

Visually Weighted Regression.

Description

Usage

Arguments

Details

Value

See Also

Examples