densityBy function

Create a 'violin plot' or density plot of the distribution of a set of variables

Create a 'violin plot' or density plot of the distribution of a set of variables

Among the many ways to describe a data set, one is a density plot for each value of a grouping variable and another is violin plot of multiple variables. A density plot shows the density for different groups to show effect sizes. A violin plot is similar to a box plot but shows the actual distribution. Median and 25th and 75th percentile lines are added to the display. If a grouping variable is specified, violinBy will draw violin plots for each variable and for each group. Data points may be drawn as well in what is known as a "raincloud plot".

violin(x,data=NULL, var=NULL, grp=NULL, grp.name=NULL, xlab=NULL, ylab=NULL,
        main="Density plot", vertical=TRUE, dots=FALSE, rain=FALSE, jitter=.05, alpha=1,
        errors=FALSE, eyes=TRUE, adjust=1, restrict=TRUE, xlim=NULL, add=FALSE, 
        col=NULL, pch=20, scale=NULL,...) 
 
violinBy(x, var=NULL, grp=NULL, data=NULL, grp.name=NULL, xlab=NULL, ylab=NULL,
        main="Density plot", vertical=TRUE, dots=FALSE, rain=FALSE, jitter=.05, alpha= 1,
        errors=FALSE, eyes=TRUE, adjust=1, restrict=TRUE, xlim=NULL, add=FALSE, 
        col=NULL, pch=20, scale=NULL,...) 
 
densityBy(x, var=NULL, grp=NULL,data=NULL, freq=FALSE, col=c("blue","red","black"), 
        alpha=.5, adjust=1, ylim=NULL, xlim=NULL, xlab="Variable", ylab="Density",
     main="Density Plot",legend=NULL)

Arguments

  • x: A matrix or data.frame (can be expressed in formula input)
  • var: The variable(s) to display
  • grp: The grouping variable(s)
  • data: The name of the data object if using formula input
  • grp.name: If the grouping variable is specified, then what names should be give to the group? Defaults to 1:ngrp
  • ylab: The y label
  • xlab: The x label
  • main: Figure title
  • vertical: If TRUE, plot the violins vertically, otherwise, horizontonally
  • dots: if TRUE, add a stripchart with the data points
  • rain: If TRUE, draw a half violin with rain drops
  • jitter: If doing a stripchart, then jitter the points this much
  • errors: If TRUE, add error bars or cats eyes to the violins
  • eyes: if TRUE and errors=TRUE, then draw cats eyes
  • alpha: A degree of transparency (0=transparent ... 1 not transparent)
  • adjust: Allows smoothing of density histograms when plotting variables like height
  • freq: if TRUE, then plot frequencies (n * density)
  • restrict: Restrict the density to the observed max and min of the data
  • xlim: if not specified, will be .5 beyond the number of variables
  • ylim: If not specified, determined by the data
  • add: Allows overplotting
  • col: Allows for specification of colours. The default for 2 groups is blue and red, for more group levels, rainbows.
  • pch: The plot character for the mean is by default a small filled circle. To not show the mean, use pch=NA
  • scale: If NULL, scale the widths by the square root of sample size, otherwise scale by the value supplied.
  • legend: If not NULL, draw a legend at c(topleft,topright,top,left,right)
  • ...: Other graphic parameters

Details

Describe the data using a violin plot. Change alpha to modify the shading. The grp variable may be used to draw separate violin plots for each of multiple groups.

For relatively smallish data sets (< 500-1000), it is informative to also show the actual data points. This done with the dots=TRUE option. The jitter value is arbitrarily set to .05, but making it larger (say .1 or .2) will display more points.

Perhaps even prettier, is draw "raincloud" plots (half violins with rain drops)

Returns

The density (y axis) by value (x axis) of the data (for densityBy) or a violin plot for each variable (perhaps broken down by groups)

Author(s)

William Revelle

Note

Formula input added July 12, 2020

See Also

describe, describeBy and statsBy for descriptive statistics and error.bars, error.bars.by and bi.bars, histBy and scatterHist for other graphic displays

Examples

violin(bfi[4:8])
violin(SATV + SATQ ~ gender, data=sat.act, grp.name =cs(MV,FV,MQ,FQ)) #formula input
violinBy(bfi,var=4:7,grp ="gender",grp.name=c("M","F"))
#rain does not work for multiple DVS, just up to 2 IVs 
violinBy(SATV  ~ education,data =sat.act, rain=TRUE, pch=".", vertical=FALSE)  #rain 

densityBy(SATV  ~ gender,data =sat.act,legend=1)  #formula input
psych packageRead PDF manual