A scatter diagram is used to visually describe the covariation between two variables

 kassambara |   17/11/2017 |   103423  |  Comments (3)  |  R Graphics Essentials

Scatter plots are used to display the relationship between two continuous variables x and y. In this article, we’ll start by showing how to create beautiful scatter plots in R.

We’ll use helper functions in the ggpubr R package to display automatically the correlation coefficient and the significance level on the plot.

We’ll also describe how to color points by groups and to add concentration ellipses around each group. Additionally, we’ll show how to create bubble charts, as well as, how to add marginal plots (histogram, density or box plot) to a scatter plot.

We continue by showing show some alternatives to the standard scatter plots, including rectangular binning, hexagonal binning and 2d density estimation. These plot types are useful in a situation where you have a large data set containing thousands of records.

R codes for zooming, in a scatter plot, are also provided. Finally, you’ll learn how to add fitted regression trend lines and equations to a scatter graph.

Contents:

1. Install cowplot package. Used to arrange multiple plots. Will be used here to create a scatter plot with marginal density plots. Install the latest developmental version as follow:

1. Install ggpmisc for adding the equation of a fitted regression line on a scatter plot:

1. Load required packages and set ggplot themes:

• Load ggplot2 and ggpubr R packages
• Set the default theme to theme_minimal() [in ggplot2]

Dataset: mtcars. The variable cyl is used as grouping variable.

Key functions:

• geom_point(): Create scatter plots. Key arguments: color, size and shape to change point color, size and shape.
• geom_smooth(): Add smoothed conditional means / regression line. Key arguments:
  • color, size and linetype: Change the line color, size and type.
  • fill: Change the fill color of the confidence region.

To remove the confidence region around the regression line, specify the argument se = FALSE in the function geom_smooth().

Change the point shape, by specifying the argument shape, for example:

To see the different point shapes commonly used in R, type this:

Create easily a scatter plot using ggscatter() [in ggpubr]. Use stat_cor() [ggpubr] to add the correlation coefficient and the significance level.

• Change point colors and shapes by groups.
• Add marginal rug: geom_rug().

• Split the plot into multiple panels. Use the function facet_wrap():

• Add concentration ellipse around groups. R function stat_ellipse(). Key arguments:
  • type: The type of ellipse. The default “t” assumes a multivariate t-distribution, and “norm” assumes a multivariate normal distribution. “euclid” draws a circle with the radius equal to level, representing the euclidean distance from the center.
  • level: The confidence level at which to draw an ellipse (default is 0.95), or, if type=“euclid”, the radius of the circle to be drawn.

Instead of drawing the concentration ellipse, you can: i) plot a convex hull of a set of points; ii) add the mean points and the confidence ellipse of each group. Key R functions: stat_chull(), stat_conf_ellipse() and stat_mean() [in ggpubr]:

Key functions:

• geom_text() and geom_label(): ggplot2 standard functions to add text to a plot.
• geom_text_repel() and geom_label_repel() [in ggrepel package]. Repulsive textual annotations. Avoid text overlapping.

First install ggrepel (ìnstall.packages("ggrepel")), then type this:

In a bubble chart, points size is controlled by a continuous variable, here qsec. In the R code below, the argument alpha is used to control color transparency. alpha should be between 0 and 1.

• Color points according to the values of the continuous variable: “mpg”.
• Change the default blue gradient color using the function scale_color_gradientn() [in ggplot2], by specifying two or more colors.

The function ggMarginal() [in ggExtra package] (Attali 2017), can be used to easily add a marginal histogram, density or box plot to a scatter plot.

First, install the ggExtra package as follow: install.packages("ggExtra"); then type the following R code:

One limitation of ggExtra is that it can’t cope with multiple groups in the scatter plot and the marginal plots.

A solution is provided in the function ggscatterhist() [ggpubr]:

In this section, we’ll present some alternatives to the standard scatter plots. These include:

• Rectangular binning. Rectangular heatmap of 2d bin counts
• Hexagonal binning: Hexagonal heatmap of 2d bin counts.
• 2d density estimation

Rectangular binning is a very useful alternative to the standard scatter plot in a situation where you have a large data set containing thousands of records.

Rectangular binning helps to handle overplotting. Rather than plotting each point, which would appear highly dense, it divides the plane into rectangles, counts the number of cases in each rectangle, and then plots a heatmap of 2d bin counts. In this plot, many small hexagon are drawn with a color intensity corresponding to the number of cases in that bin.

Key function: geom_bin2d(): Creates a heatmap of 2d bin counts. Key arguments: bins, numeric vector giving number of bins in both vertical and horizontal directions. Set to 30 by default.

1. Hexagonal binning: Similar to rectangular binning, but divides the plane into regular hexagons. Hexagon bins avoid the visual artefacts sometimes generated by the very regular alignment of `geom_bin2d().

Key function: geom_hex()

1. Contours of a 2d density estimate. Perform a 2D kernel density estimation and display results as contours overlaid on the scatter plot. This can be also useful for dealing with overplotting.

Key function: geom_density_2d()

• Create a scatter plot with rectangular and hexagonal binning:

• Create a scatter plot with 2d density estimation:

• Key function: facet_zomm() [in ggforce] (Pedersen 2016).
• Demo data set: iris. The R code below zoom the points where Species == "versicolor".

To zoom the points, where Petal.Length < 2.5, type this:

In this section, we’ll describe how to add trend lines to a scatter plot and labels (equation, R2, BIC, AIC) for a fitted lineal model.

1. Load packages and create a basic scatter plot facetted by groups:

1. Add regression line, correlation coefficient and equantions of the fitted line. Key functions:
   • stat_smooth() [ggplot2]
   • stat_cor() [ggpubr]
   • stat_poly_eq()[ggpmisc]

• Fit polynomial regression line and add labels:

Note that, you can also display the AIC and the BIC values using ..AIC.label.. and ..BIC.label.. in the above equation.

Other arguments (label.x, label.y) are available in the function stat_poly_eq() to adjust label positions.

For more examples, type this R code: browseVignettes(“ggpmisc”).

1. Create a basic scatter plot:

Possible layers, include:

• geom_point() for scatter plot
• geom_smooth() for adding smoothed line such as regression line
• geom_rug() for adding a marginal rug
• geom_text() for adding textual annotations

1. Continuous bivariate distribution:

Possible layers include:

• geom_bin2d(): Rectangular binning.
• geom_hex(): Hexagonal binning.
• geom_density_2d(): Contours from a 2d density estimate

Enjoyed this article? Give us 5 stars (just above this text block)! Reader needs to be STHDA member for voting. I’d be very grateful if you’d help it spread by emailing it to a friend, or sharing it on Twitter, Facebook or Linked In.

Show me some love with the like buttons below... Thank you and please don't forget to share and comment below!!

Avez vous aimé cet article? Donnez nous 5 étoiles (juste au dessus de ce block)! Vous devez être membre pour voter. Je vous serais très reconnaissant si vous aidiez à sa diffusion en l'envoyant par courriel à un ami ou en le partageant sur Twitter, Facebook ou Linked In.

Montrez-moi un peu d'amour avec les like ci-dessous ... Merci et n'oubliez pas, s'il vous plaît, de partager et de commenter ci-dessous!