Graphics and Data Visualization in R

318 minute read

Source code downloads:     [ .Rmd ]     [ .R ]

Overview

Graphics in R

  • Powerful environment for visualizing scientific data
  • Integrated graphics and statistics infrastructure
  • Publication quality graphics
  • Fully programmable
  • Highly reproducible
  • Full LaTeX, Sweave, knitr and R Markdown support.
  • Vast number of R packages with graphics utilities

Documentation on Graphics in R

Graphics Environments

  • Viewing and savings graphics in R
    • On-screen graphics
    • postscript, pdf, svg
    • jpeg/png/wmf/tiff/…
  • Four major graphics environments

Base Graphics

Overview

  • Important high-level plotting functions
    • plot: generic x-y plotting
    • barplot: bar plots
    • boxplot: box-and-whisker plot
    • hist: histograms
    • pie: pie charts
    • dotchart: cleveland dot plots
    • image, heatmap, contour, persp: functions to generate image-like plots
    • qqnorm, qqline, qqplot: distribution comparison plots
    • pairs, coplot: display of multivariant data
  • Help on these functions
    • ?myfct
    • ?plot
    • ?par

Preferred Input Data Objects

  • Matrices and data frames
  • Vectors
  • Named vectors

Scatter Plots

Basic scatter plots

Sample data set for subsequent plots

set.seed(1410)
y <- matrix(runif(30), ncol=3, dimnames=list(letters[1:10], LETTERS[1:3]))
y
##            A          B            C
## a 0.26904539 0.47439030 0.4427788756
## b 0.53178658 0.31128960 0.3233293493
## c 0.93379571 0.04576263 0.0004628517
## d 0.14314802 0.12066723 0.4104402000
## e 0.57627063 0.83251909 0.9884746270
## f 0.49001235 0.38298651 0.8235850153
## g 0.66562596 0.70857731 0.7490944304
## h 0.50089252 0.24772695 0.2117313873
## i 0.57033245 0.06044799 0.8776291364
## j 0.04087422 0.85814118 0.1061618729
plot(y[,1], y[,2]) 

All pairs

pairs(y) 

Plot labels

plot(y[,1], y[,2], pch=20, col="red", main="Symbols and Labels")
text(y[,1]+0.03, y[,2], rownames(y))

More examples

Print instead of symbols the row names

plot(y[,1], y[,2], type="n", main="Plot of Labels")
text(y[,1], y[,2], rownames(y)) 

Usage of important plotting parameters

grid(5, 5, lwd = 2) 
op <- par(mar=c(8,8,8,8), bg="lightblue")
plot(y[,1], y[,2], type="p", col="red", cex.lab=1.2, cex.axis=1.2, 
     cex.main=1.2, cex.sub=1, lwd=4, pch=20, xlab="x label", 
     ylab="y label", main="My Main", sub="My Sub")
par(op)

Important arguments}

  • mar: specifies the margin sizes around the plotting area in order: c(bottom, left, top, right)
  • col: color of symbols
  • pch: type of symbols, samples: example(points)
  • lwd: size of symbols
  • cex.*: control font sizes
  • For details see ?par

Add a regression line to a plot

plot(y[,1], y[,2])
myline <- lm(y[,2]~y[,1]); abline(myline, lwd=2) 
summary(myline) 
## 
## Call:
## lm(formula = y[, 2] ~ y[, 1])
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -0.40357 -0.17912 -0.04299  0.22147  0.46623 
## 
## Coefficients:
##             Estimate Std. Error t value Pr(>|t|)  
## (Intercept)   0.5764     0.2110   2.732   0.0258 *
## y[, 1]       -0.3647     0.3959  -0.921   0.3839  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.3095 on 8 degrees of freedom
## Multiple R-squared:  0.09589,    Adjusted R-squared:  -0.01712 
## F-statistic: 0.8485 on 1 and 8 DF,  p-value: 0.3839

Same plot as above, but on log scale

plot(y[,1], y[,2], log="xy") 

Add a mathematical expression to a plot

plot(y[,1], y[,2]); text(y[1,1], y[1,2], 
     expression(sum(frac(1,sqrt(x^2*pi)))), cex=1.3) 

Exercise 1

  • Task 1: Generate scatter plot for first two columns in iris data frame and color dots by its Species column.
  • Task 2: Use the xlim/ylim arguments to set limits on the x- and y-axes so that all data points are restricted to the left bottom quadrant of the plot.

Structure of iris data set:

class(iris)
## [1] "data.frame"
iris[1:4,]
##   Sepal.Length Sepal.Width Petal.Length Petal.Width Species
## 1          5.1         3.5          1.4         0.2  setosa
## 2          4.9         3.0          1.4         0.2  setosa
## 3          4.7         3.2          1.3         0.2  setosa
## 4          4.6         3.1          1.5         0.2  setosa
table(iris$Species)
## 
##     setosa versicolor  virginica 
##         50         50         50

Line Plots

Single Data Set

plot(y[,1], type="l", lwd=2, col="blue") 

Additional lines can be added to an existing plot with the lines() function.

plot(y[,1], type="l", lwd=2, col="blue")
lines(y[,2], lwd=2, lty=1,  col="red")
legend(8.3, 0.95, legend=c("Line 1", "Line 2"), col=c("blue", "red"), lty=1)

Many Data Sets

Alterntively, one can plot a line graph for all columns in data frame y with the following approach. The split.screen function is used in this example in a for loop to overlay several line graphs in the same plot.

split.screen(c(1,1)) 
## [1] 1
plot(y[,1], ylim=c(0,1), xlab="Measurement", ylab="Intensity", type="l", lwd=2, col=1)
for(i in 2:length(y[1,])) { 
    screen(1, new=FALSE)
    plot(y[,i], ylim=c(0,1), type="l", lwd=2, col=i, xaxt="n", yaxt="n", ylab="", 
             xlab="", main="", bty="n") 
}
close.screen(all=TRUE) 

Bar Plots

Basics

barplot(y[1:4,], ylim=c(0, max(y[1:4,])+0.3), beside=TRUE, 
        legend=letters[1:4]) 
text(labels=round(as.vector(as.matrix(y[1:4,])),2), x=seq(1.5, 13, by=1)
     +sort(rep(c(0,1,2), 4)), y=as.vector(as.matrix(y[1:4,]))+0.04) 

Error bars

bar <- barplot(m <- rowMeans(y) * 10, ylim=c(0, 10))
stdev <- sd(t(y))
arrows(bar, m, bar, m + stdev, length=0.15, angle = 90)

Mirrored bar plot

df <- data.frame(group = rep(c("Above", "Below"), each=10), x = rep(1:10, 2), y = c(runif(10, 0, 1), runif(10, -1, 0)))
plot(c(0,12), range(df$y), type = "n")
barplot(height = df$y[df$group == "Above"], add = TRUE, axes = FALSE)
barplot(height = df$y[df$group == "Below"], add = TRUE, axes = FALSE)

Bar plot of loan payments and amortization tables

The following imports a mortgage payment function (from here) that calculates monthly and annual mortgage/loan payments, generates amortization tables and plots the results in form of a bar plot. A Shiny App using this function has been created by Antoine Soetewey here.

source("https://raw.githubusercontent.com/tgirke/GEN242/main/content/en/tutorials/rgraphics/scripts/mortgage.R")
## The monthly mortgage payments and amortization rates can be calculted with the mortgage() function like this: 
##  
##     m <- mortgage(P=500000, I=6, L=30, plotData=TRUE)
##         P = principal (loan amount)
##         I = annual interest rate
##         L = length of the loan in years
m <- mortgage(P=250000, I=6, L=15, plotData=TRUE)
## 
## The payments for this loan are:
##  
##             Monthly payment: $2109.642 (stored in m$monthPay)
## 
##             Total cost: $379735.6
## 
## The amortization data for each of the 180 months are stored in "m$aDFmonth".
## 
## The amortization data for each of the 15 years are stored in "m$aDFyear".

Histograms

hist(y, freq=TRUE, breaks=10)

Density Plots}

plot(density(y), col="red")

Pie Charts

pie(y[,1], col=rainbow(length(y[,1]), start=0.1, end=0.8), clockwise=TRUE)
legend("topright", legend=row.names(y), cex=1.3, bty="n", pch=15, pt.cex=1.8, 
col=rainbow(length(y[,1]), start=0.1, end=0.8), ncol=1) 

Color Selection Utilities

Default color palette and how to change it

palette()
## [1] "black"   "#DF536B" "#61D04F" "#2297E6" "#28E2E5" "#CD0BBC" "#F5C710" "gray62"
palette(rainbow(5, start=0.1, end=0.2))
palette()
## [1] "#FF9900" "#FFBF00" "#FFE600" "#F2FF00" "#CCFF00"
palette("default")

The gray function allows to select any type of gray shades by providing values from 0 to 1

gray(seq(0.1, 1, by= 0.2))
## [1] "#1A1A1A" "#4D4D4D" "#808080" "#B3B3B3" "#E6E6E6"

Color gradients with colorpanel function from gplots library

library(gplots)
colorpanel(5, "darkblue", "yellow", "white")

Much more on colors in R see Earl Glynn’s color chart

Arranging Several Plots on Single Page

With par(mfrow=c(nrow, ncol)) one can define how several plots are arranged next to each other.

par(mfrow=c(2,3)) 
for(i in 1:6) plot(1:10) 

Arranging Plots with Variable Width

The layout function allows to divide the plotting device into variable numbers of rows and columns with the column-widths and the row-heights specified in the respective arguments.

nf <- layout(matrix(c(1,2,3,3), 2, 2, byrow=TRUE), c(3,7), c(5,5), 
             respect=TRUE)
# layout.show(nf)
for(i in 1:3) barplot(1:10) 

Saving Graphics to Files

After the pdf() command all graphs are redirected to file test.pdf. Works for all common formats similarly: jpeg, png, ps, tiff, …

pdf("test.pdf"); plot(1:10, 1:10); dev.off() 

Generates Scalable Vector Graphics (SVG) files that can be edited in vector graphics programs, such as InkScape.

svg("test.svg"); plot(1:10, 1:10); dev.off() 

Exercise 2

Bar plots

  • Task 1: Calculate the mean values for the Species components of the first four columns in the iris data set. Organize the results in a matrix where the row names are the unique values from the iris Species column and the column names are the same as in the first four iris columns.
  • Task 2: Generate two bar plots: one with stacked bars and one with horizontally arranged bars.

Structure of iris data set:

class(iris)
## [1] "data.frame"
iris[1:4,]
##   Sepal.Length Sepal.Width Petal.Length Petal.Width Species
## 1          5.1         3.5          1.4         0.2  setosa
## 2          4.9         3.0          1.4         0.2  setosa
## 3          4.7         3.2          1.3         0.2  setosa
## 4          4.6         3.1          1.5         0.2  setosa
table(iris$Species)
## 
##     setosa versicolor  virginica 
##         50         50         50

Grid Graphics

  • What is grid?
    • Low-level graphics system
    • Highly flexible and controllable system
    • Does not provide high-level functions
    • Intended as development environment for custom plotting functions
    • Pre-installed on new R distributions
  • Documentation and Help

lattice Graphics

  • What is lattice?
    • High-level graphics system
    • Developed by Deepayan Sarkar
    • Implements Trellis graphics system from S-Plus
    • Simplifies high-level plotting tasks: arranging complex graphical features
    • Syntax similar to R’s base graphics
  • Documentation and Help

Open a list of all functions available in the lattice package

library(lattice) 
library(help=lattice) 

Accessing and changing global parameters:

?lattice.options
?trellis.device

Scatter Plot Sample

library(lattice)
p1 <- xyplot(1:8 ~ 1:8 | rep(LETTERS[1:4], each=2), as.table=TRUE) 
plot(p1)

Line Plot Sample

library(lattice)
p2 <- parallelplot(~iris[1:4] | Species, iris, horizontal.axis = FALSE, 
              layout = c(1, 3, 1))  
plot(p2)

ggplot2 Graphics

  • What is ggplot2?
    • High-level graphics system developed by Hadley Wickham
    • Implements grammar of graphics from Leland Wilkinson
    • Streamlines many graphics workflows for complex plots
    • Syntax centered around main ggplot function
    • Simpler qplot function provides many shortcuts
  • Documentation and Help

Design Concept of ggplot2

Plotting formalized and implemented by the grammar of graphics by Leland Wilkinson and Hadley Wickham (Wickham 2010, 2009; Wilkinson 2012). The plotting process in ggplot2 is devided into layers including:

  1. Data: the actual data to be plotted
  2. Aesthetics: the scales onto which the data will be mapped
  3. Geometries: shapes used to represent data (e.g. bar or scatter plot)
  4. Facets: row and column layout of sub-plots
  5. Statistics: data models and summaries
  6. Coordinates: the plotting space
  7. Theme: styles to be used, such as fonts, backgrounds, etc.

ggplot2 Usage

  • ggplot function accepts two main arguments
    • Data set to be plotted
    • Aesthetic mappings provided by aes function
  • Additional parameters such as geometric objects (e.g. points, lines, bars) are passed on by appending them with + as separator.
  • List of available geom_* functions see here
  • Settings of plotting theme can be accessed with the command theme_get() and its settings can be changed with theme().
  • Preferred input data object
    • qplot: data.frame or tibble (support for vector, matrix, ...)
    • ggplot: data.frame or tibble
  • Packages with convenience utilities to create expected inputs
    • dplyr (plyr)
    • tidyr and reshape2

qplot Function

The syntax of qplot is similar as R’s basic plot function

  • Arguments
    • x: x-coordinates (e.g. col1)
    • y: y-coordinates (e.g. col2)
    • data: data.frame or tibble with corresponding column names
    • xlim, ylim: e.g. xlim=c(0,10)
    • log: e.g. log="x" or log="xy"
    • main: main title; see ?plotmath for mathematical formula
    • xlab, ylab: labels for the x- and y-axes
    • color, shape, size
    • ...: many arguments accepted by plot function

qplot: scatter plot basics

Create sample data, here 3 vectors: x, y and cat

library(ggplot2)
x <- sample(1:10, 10); y <- sample(1:10, 10); cat <- rep(c("A", "B"), 5)

Simple scatter plot

qplot(x, y, geom="point")

Prints dots with different sizes and colors

qplot(x, y, geom="point", size=x, color=cat, 
      main="Dot Size and Color Relative to Some Values")

Drops legend

qplot(x, y, geom="point", size=x, color=cat) + 
      theme(legend.position = "none")

Plot different shapes

qplot(x, y, geom="point", size=5, shape=cat)

Colored groups

p <- qplot(x, y, geom="point", size=x, color=cat, 
            main="Dot Size and Color Relative to Some Values") + 
     theme(legend.position = "none")
print(p)

Regression line

set.seed(1410)
dsmall <- diamonds[sample(nrow(diamonds), 1000), ]
p <- qplot(carat, price, data = dsmall) +
           geom_smooth(method="lm")
print(p)

Local regression curve (loess)

p <- qplot(carat, price, data=dsmall, geom=c("point", "smooth")) 
print(p) # Setting se=FALSE removes error shade

ggplot Function

  • More important than qplot to access full functionality of ggplot2
  • Main arguments
    • data set, usually a data.frame or tibble
    • aesthetic mappings provided by aes function
  • General ggplot syntax
    • ggplot(data, aes(...)) + geom() + ... + stat() + ...
  • Layer specifications
    • geom(mapping, data, ..., geom, position)
    • stat(mapping, data, ..., stat, position)
  • Additional components
    • scales
    • coordinates
    • facet
  • aes() mappings can be passed on to all components (ggplot, geom, etc.). Effects are global when passed on to ggplot() and local for other components.
    • x, y
    • color: grouping vector (factor)
    • group: grouping vector (factor)

Changing Plotting Themes in ggplot

  • Theme settings can be accessed with theme_get()
  • Their settings can be changed with theme()

Example how to change background color to white

... + theme(panel.background=element_rect(fill = "white", colour = "black")) 

Storing ggplot Specifications

Plots and layers can be stored in variables

p <- ggplot(dsmall, aes(carat, price)) + geom_point() 
p # or print(p)

Returns information about data and aesthetic mappings followed by each layer

summary(p) 

Print dots with different sizes and colors

bestfit <- geom_smooth(method = "lm", se = F, color = alpha("steelblue", 0.5), size = 2)
p + bestfit # Plot with custom regression line

Syntax to pass on other data sets

p %+% diamonds[sample(nrow(diamonds), 100),] 

Saves plot stored in variable p to file

ggsave(p, file="myplot.pdf") 

Standard R export functons for graphics work as well (see here).

ggplot: scatter plots

Basic example

set.seed(1410)  
dsmall <- as.data.frame(diamonds[sample(nrow(diamonds), 1000), ])
p <- ggplot(dsmall, aes(carat, price, color=color)) + 
            geom_point(size=4)
print(p) 

Interactive version of above plot can be generated with the ggplotly function from the plotly package.

library(plotly)
ggplotly(p)

Regression line

p <- ggplot(dsmall, aes(carat, price)) + geom_point() + 
            geom_smooth(method="lm", se=FALSE) +
            theme(panel.background=element_rect(fill = "white", colour = "black"))
print(p) 

Several regression lines

p <- ggplot(dsmall, aes(carat, price, group=color)) + 
            geom_point(aes(color=color), size=2) + 
            geom_smooth(aes(color=color), method = "lm", se=FALSE) 
print(p) 

Local regression curve (loess)

p <- ggplot(dsmall, aes(carat, price)) + geom_point() + geom_smooth() 
print(p) # Setting se=FALSE removes error shade

ggplot: line plot

p <- ggplot(iris, aes(Petal.Length, Petal.Width, group=Species, 
            color=Species)) + geom_line() 
print(p) 

Faceting

p <- ggplot(iris, aes(Sepal.Length, Sepal.Width)) + 
            geom_line(aes(color=Species), size=1) + 
            facet_wrap(~Species, ncol=1)
print(p) 

Exercise 3

Scatter plots with ggplot2

  • Task 1: Generate scatter plot for first two columns in iris data frame and color dots by its Species column.
  • Task 2: Use the xlim and ylim arguments to set limits on the x- and y-axes so that all data points are restricted to the left bottom quadrant of the plot.
  • Task 3: Generate corresponding line plot with faceting presenting the individual data sets in saparate plots.

Structure of iris data set

class(iris)
## [1] "data.frame"
iris[1:4,]
##   Sepal.Length Sepal.Width Petal.Length Petal.Width Species
## 1          5.1         3.5          1.4         0.2  setosa
## 2          4.9         3.0          1.4         0.2  setosa
## 3          4.7         3.2          1.3         0.2  setosa
## 4          4.6         3.1          1.5         0.2  setosa
table(iris$Species)
## 
##     setosa versicolor  virginica 
##         50         50         50

Bar Plots

Sample Set: the following transforms the iris data set into a ggplot2-friendly format.

Calculate mean values for aggregates given by Species column in iris data set

iris_mean <- aggregate(iris[,1:4], by=list(Species=iris$Species), FUN=mean) 

Calculate standard deviations for aggregates given by Species column in iris data set

iris_sd <- aggregate(iris[,1:4], by=list(Species=iris$Species), FUN=sd) 

Reformat iris_mean with melt from wide to long form as expected by ggplot2. Newer alternatives for restructuring data.frames and tibbles from wide into long form use the gather and pivot_longer functions defined by the tidyr package. Their usage is shown below as well. The functions pivot_longer and pivot_wider are expected to provide the most flexible long-term solution, but may not work in older R versions.

library(reshape2) # Defines melt function
df_mean <- melt(iris_mean, id.vars=c("Species"), variable.name = "Samples", value.name="Values")
df_mean2 <- tidyr::gather(iris_mean, !Species, key = "Samples", value = "Values")
df_mean3 <- tidyr::pivot_longer(iris_mean, !Species, names_to="Samples", values_to="Values") 

Reformat iris_sd with melt

df_sd <- melt(iris_sd, id.vars=c("Species"), variable.name = "Samples", value.name="Values")

Define standard deviation limits

limits <- aes(ymax = df_mean[,"Values"] + df_sd[,"Values"], ymin=df_mean[,"Values"] - df_sd[,"Values"])

Verical orientation

p <- ggplot(df_mean, aes(Samples, Values, fill = Species)) + 
            geom_bar(position="dodge", stat="identity")
print(p) 

To enforce that the bars are plotted in the order specified in the input data, one can instruct ggplot to do so by turning the corresponding column (here Species) into an ordered factor as follows.

df_mean$Species <- factor(df_mean$Species, levels=unique(df_mean$Species), ordered=TRUE)

In the above example this is not necessary since ggplot uses this order already.

Horizontal orientation

p <- ggplot(df_mean, aes(Samples, Values, fill = Species)) + 
            geom_bar(position="dodge", stat="identity") + coord_flip() + 
            theme(axis.text.y=element_text(angle=0, hjust=1))
print(p) 

Faceting

p <- ggplot(df_mean, aes(Samples, Values)) + geom_bar(aes(fill = Species), stat="identity") +
            facet_wrap(~Species, ncol=1)
print(p) 

Error bars

p <- ggplot(df_mean, aes(Samples, Values, fill = Species)) + 
            geom_bar(position="dodge", stat="identity") + 
            geom_errorbar(limits, position="dodge") 
print(p) 

Mirrored

df <- data.frame(group = rep(c("Above", "Below"), each=10), x = rep(1:10, 2), y = c(runif(10, 0, 1), runif(10, -1, 0)))
p <- ggplot(df, aes(x=x, y=y, fill=group)) + 
            geom_col()
print(p) 

Changing Color Settings

library(RColorBrewer)
# display.brewer.all() 
p <- ggplot(df_mean, aes(Samples, Values, fill=Species, color=Species)) +
            geom_bar(position="dodge", stat="identity") + geom_errorbar(limits, position="dodge") + 
            scale_fill_brewer(palette="Blues") + scale_color_brewer(palette = "Greys") 
print(p) 

Using standard R color theme

p <- ggplot(df_mean, aes(Samples, Values, fill=Species, color=Species)) + 
            geom_bar(position="dodge", stat="identity") + geom_errorbar(limits, position="dodge") + 
            scale_fill_manual(values=c("red", "green3", "blue")) + 
            scale_color_manual(values=c("red", "green3", "blue")) 
print(p) 

Exercise 4

Bar plots

  • Task 1: Calculate the mean values for the Species components of the first four columns in the iris data set. Use the melt function from the reshape2 package to bring the data into the expected format for ggplot.
  • Task 2: Generate two bar plots: one with stacked bars and one with horizontally arranged bars.

Structure of iris data set

class(iris)
## [1] "data.frame"
iris[1:4,]
##   Sepal.Length Sepal.Width Petal.Length Petal.Width Species
## 1          5.1         3.5          1.4         0.2  setosa
## 2          4.9         3.0          1.4         0.2  setosa
## 3          4.7         3.2          1.3         0.2  setosa
## 4          4.6         3.1          1.5         0.2  setosa
table(iris$Species)
## 
##     setosa versicolor  virginica 
##         50         50         50

Data reformatting example

Here for line plot

y <- matrix(rnorm(500), 100, 5, dimnames=list(paste("g", 1:100, sep=""), paste("Sample", 1:5, sep="")))
y <- data.frame(Position=1:length(y[,1]), y)
y[1:4, ] # First rows of input format expected by melt()
##    Position    Sample1    Sample2    Sample3    Sample4     Sample5
## g1        1  1.5336975 -1.0365027 -2.0276195 -0.4580396 -0.06460952
## g2        2 -2.0960304  2.1878704  0.7260334  0.8274617  0.24192162
## g3        3 -0.8233125  0.4250477  0.6526331 -0.4509962 -1.06778801
## g4        4  1.0961555  0.8101104 -0.3403762 -0.7222191 -0.72737741
df <- melt(y, id.vars=c("Position"), variable.name = "Samples", value.name="Values")
p <- ggplot(df, aes(Position, Values)) + geom_line(aes(color=Samples)) + facet_wrap(~Samples, ncol=1)
print(p)

Same data can be represented in box plot as follows

ggplot(df, aes(Samples, Values, fill=Samples)) + geom_boxplot() + geom_jitter(color="darkgrey")

Jitter Plots

p <- ggplot(dsmall, aes(color, price/carat)) + 
            geom_jitter(alpha = I(1 / 2), aes(color=color))
print(p) 

Box plots

p <- ggplot(dsmall, aes(color, price/carat, fill=color)) + geom_boxplot()
print(p) 

Violin plots

p <- ggplot(dsmall, aes(color, price/carat, fill=color)) + geom_violin()                                                                                                      
print(p)                                                                                                                                                                      

Same violin plot as interactive plot generated with ggplotly, where the actual data points are shown as well by including geom_jitter().

p <- ggplot(dsmall, aes(color, price/carat, fill=color)) + geom_violin() + geom_jitter(aes(color=color))
ggplotly(p)