Amit Kohli Amit Kohli - 23 days ago 5
R Question

buffer areas around lines ggplot2

I would like a chart that as a function of the day in the year, advances from 0 -> 100% in both x and y axes (where each axis is a separate metric). Depending on where the data is relative to the day in the year, I'd like to show whether that's good or bad. Quite simply, I can show it like this:
enter image description here

So the above plot shows we're in a good sitaution because the "tip" (darkest biggest points) are past the 50% mark (and assume we are 50% through the year). But I wanted to add gradient lines around the horizontal and vertical lines to show more nuance. Here's an explanation of the areas (first drawing is the explanation... and the second one is the way I would like to show this in ggplot... with the area fully filled-in.

enter image description here

This is how far I have come in ggplot:

enter image description here

Problems I'm having:


  1. For some reason, the vertical gradient isn't accepting the alpha parameter

  2. I can't assign two different gradients, once I define the gradient, it applies to the vertical and horizontal one.

  3. This looks terrible. Is there a better approach I should be following?



Are Problems 1-2 solvable? If anyone has a better approach not using
geom_line
, please feel free to suggest approach.

EDIT: As the lines move, so would the gradients, so a static background wouldn't work here.

Code follows:

dff <- data.frame(x = 1:60+(runif(n = 60,-2,2)),
y = 1:60+(runif(n = 60,-2,2)),
z = 1:60)

dfgrad <- data.frame(static = c(rep(50,1000)), line = seq(0,100,length.out=100))

## To see the gradientlines thinner, change the size on the geom_line to like 200

ggplot(dff,aes(x,y)) +
geom_line(data = dfgrad, aes(x=static, y=line, color=line),size=1000,alpha=0.5) +
geom_line(data = dfgrad, aes(x=line, y=static, color=line),size=1000,alpha=0.5) +
scale_colour_gradientn( colours = c( "yellow", "darkgreen","darkred"),
breaks = c( 0, 3, 100),
limits = c( 0,100)) +
geom_hline(yintercept = 50, linetype="dashed") +
geom_vline(xintercept = 50, linetype="dashed") +
geom_point(aes(alpha=dff$z,size= (dff$z))) +
theme(legend.position="none") +
scale_x_continuous(expand = c(0, 0)) + scale_y_continuous(expand = c(0, 0))

Answer

If I were you, I'd make rectangles by grid package and put them on the graph using annotation_custom(). (your problem.1 is due to overlaying, try alpha=0.05)

Here is my example:

library(ggplot2); library(grid); library(dplyr)

g1 <- colorRampPalette(c("yellow", "darkgreen","darkred"))(20) %>%
  alpha(0.5) %>% matrix(ncol=1) %>%  # up and down gradient
  rasterGrob(width = 1, height = 1)  # full-size (control it by ggplot2)

g2 <- colorRampPalette(c("cyan", "darkgreen","darkblue"))(20) %>% 
  alpha(0.5) %>% matrix(nrow=1) %>%  # left and right gradient
  rasterGrob(width = 1, height = 1)

ggplot(dff,aes(x,y)) +
  annotation_custom(g1, xmin = 35, xmax = 65, ymin = -3, ymax = 100) + 
  annotation_custom(g2, xmin = -3, xmax = 100, ymin = 35, ymax = 65) + 
  geom_hline(yintercept = 50, linetype="dashed") +
  geom_vline(xintercept = 50, linetype="dashed") +
  geom_point(aes(alpha=dff$z,size= (dff$z))) +
  theme(legend.position="none") +
  coord_cartesian(xlim = c(-3, 100), ylim = c(-3, 100), expand = F)

enter image description here