Multivariate Graphs

• grouping

• faceting

• bubble chart

• correlation matrix

• 3D scatter plot

• mosaic chart

Two categorical variables

ggplot(data, aes(x =, color = )) +
  geom_bar(position =)

Position

• stack
• dodge
• fill

Two quantitative variables

ggplot(data, aes(x =, y = )) +
  geom_point()

Common options

• color
• alpha
• size

ggplot(data, aes(x =, y = )) +
  geom_point() +
  geom_smooth(method = , formula = , se =)

For geom_smooth

• method: “lm”, “loess”, “gam”
• formula examples: y ~ x, y ~ poly(x, 2)
• se: with confidence intervals TRUE FALSE

ggplot(data, aes(x =, y = )) +
  geom_boxplot()

Common options

• fill
• outlier.color
• outlier.size
• notched

ggplot(data, aes(x =, y =)) +
  geom_violin()

Common options

• color
• fill

library(dplyr)
plotdata <- data %>%
  group_by(x) %>%
  summarize(n = n(),
            mean = mean(y),
            se = sd(y) / sqrt(y))

ggplot(plotdata, 
       aes(x = rank, y = mean, group = 1)) +
  geom_point() +
  geom_line() +
  geom_errorbar(aes(ymin = mean - se, ymax = mean + se), 
                width = )

ggplot(data, aes(x =, y =, fill = )) +
  geom_density(alpha = )

Common options

• alpha

library(ggridges)
ggplot(data, 
       aes(x = , y =, fill =)) +
  geom_density_ridges() + 
  theme_ridges() +
  theme(legend.position = "none")

• can handle many categories

ggplot(data, aes(x =, y = )) +
  geom_jitter()

• can be combined with color mapping and boxplots
• can also be used in scatter plots to avoid overprinting

library(ggbeeswarm)
ggplot(data, aes(x = , y =)) +
  geom_quasirandom() 

Common options

• alpha
• size
• can be combined with color mapping

Categorical variable with many levels vs. quantiative variable

ggplot(plotdata, aes(x=quantvar, y=reorder(catvar, quantvar))) +
  geom_point()

3 quantitative variables

ggplot(data, aes(x = , y = , size = )) +
  geom_point()

Common options

• shape (e.g. 21 with fill and color)
• alpha

+ scale_size_continuous(range = c(1, 10))

r <- cor(df, use="complete.obs")
library(ggcorrplot)
ggcorrplot(r, 
           hc.order = TRUE, # order variables into clusters? 
           type = "lower",  # upper, lower, or full matrix?
           lab = TRUE)      # include numeric labels?

2 or more categorical variables

library(vcd)
mosaic(~var1 + var2 + ..., data, shade=TRUE)

hard to read after 4 variables

static

library(scatterplot3d)
with(data, scatterplot3d(x=, y=, z=))

interactive

library(car)
with(data, scatter3d(x=, y=, z=))