Scatterplots (ggplot2)

Problem

You want to make a scatterplot.

Solution

Suppose this is your data:

  1. set.seed(955)
  2. # Make some noisily increasing data
  3. dat <- data.frame(cond = rep(c("A", "B"), each=10),
  4.                   xvar = 1:20 + rnorm(20,sd=3),
  5.                   yvar = 1:20 + rnorm(20,sd=3))
  6. head(dat)
  7. #>   cond      xvar         yvar
  8. #> 1    A -4.252354  3.473157275
  9. #> 2    A  1.702318  0.005939612
  10. #> 3    A  4.323054 -0.094252427
  11. #> 4    A  1.780628  2.072808278
  12. #> 5    A 11.537348  1.215440358
  13. #> 6    A  6.672130  3.608111411
  15. library(ggplot2)

Basic scatterplots with regression lines

  1. ggplot(dat, aes(x=xvar, y=yvar)) +
  2.     geom_point(shape=1)      # Use hollow circles
  4. ggplot(dat, aes(x=xvar, y=yvar)) +
  5.     geom_point(shape=1) +    # Use hollow circles
  6.     geom_smooth(method=lm)   # Add linear regression line 
  7.                              #  (by default includes 95% confidence region)
  9. ggplot(dat, aes(x=xvar, y=yvar)) +
  10.     geom_point(shape=1) +    # Use hollow circles
  11.     geom_smooth(method=lm,   # Add linear regression line
  12.                 se=FALSE)    # Don't add shaded confidence region
  15. ggplot(dat, aes(x=xvar, y=yvar)) +
  16.     geom_point(shape=1) +    # Use hollow circles
  17.     geom_smooth()            # Add a loess smoothed fit curve with confidence region
  18. #> `geom_smooth()` using method = 'loess'

plot of chunk unnamed-chunk-3plot of chunk unnamed-chunk-3plot of chunk unnamed-chunk-3plot of chunk unnamed-chunk-3

Set color/shape by another variable

  1. # Set color by cond
  2. ggplot(dat, aes(x=xvar, y=yvar, color=cond)) + geom_point(shape=1)
  4. # Same, but with different colors and add regression lines
  5. ggplot(dat, aes(x=xvar, y=yvar, color=cond)) +
  6.     geom_point(shape=1) +
  7.     scale_colour_hue(l=50) + # Use a slightly darker palette than normal
  8.     geom_smooth(method=lm,   # Add linear regression lines
  9.                 se=FALSE)    # Don't add shaded confidence region
  11. # Extend the regression lines beyond the domain of the data
  12. ggplot(dat, aes(x=xvar, y=yvar, color=cond)) + geom_point(shape=1) +
  13.     scale_colour_hue(l=50) + # Use a slightly darker palette than normal
  14.     geom_smooth(method=lm,   # Add linear regression lines
  15.                 se=FALSE,    # Don't add shaded confidence region
  16.                 fullrange=TRUE) # Extend regression lines
  19. # Set shape by cond
  20. ggplot(dat, aes(x=xvar, y=yvar, shape=cond)) + geom_point()
  22. # Same, but with different shapes
  23. ggplot(dat, aes(x=xvar, y=yvar, shape=cond)) + geom_point() +
  24.     scale_shape_manual(values=c(1,2))  # Use a hollow circle and triangle

plot of chunk unnamed-chunk-4plot of chunk unnamed-chunk-4plot of chunk unnamed-chunk-4plot of chunk unnamed-chunk-4plot of chunk unnamed-chunk-4

See Colors (ggplot2)) and Shapes and line types for more information about colors and shapes.

Handling overplotting

If you have many data points, or if your data scales are discrete, then the data points might overlap and it will be impossible to see if there are many points at the same location.

  1. # Round xvar and yvar to the nearest 5
  2. dat$xrnd <- round(dat$xvar/5)*5
  3. dat$yrnd <- round(dat$yvar/5)*5
  5. # Make each dot partially transparent, with 1/4 opacity
  6. # For heavy overplotting, try using smaller values
  7. ggplot(dat, aes(x=xrnd, y=yrnd)) +
  8.     geom_point(shape=19,      # Use solid circles
  9.                alpha=1/4)     # 1/4 opacity
  12. # Jitter the points
  13. # Jitter range is 1 on the x-axis, .5 on the y-axis
  14. ggplot(dat, aes(x=xrnd, y=yrnd)) +
  15.     geom_point(shape=1,      # Use hollow circles
  16.                position=position_jitter(width=1,height=.5))

plot of chunk unnamed-chunk-5plot of chunk unnamed-chunk-5