ggplot

Basic components of ggplot2 Plot

• A data frame
• aesthetic mappings: How data are mapped to color, size.
• geoms: Geometric objects like points, lines, shapes.
• facets: For conditional plots
• stats: statistical transformations like binning, quantiles, smoothing.
• scales: What scale an aesthetic map uses(example, male = red, female = blue)
• coordinate system

Making basic plot

Importing necessary libraries

invisible(library(ggplot2))
if("nycflights13" %in% rownames(installed.packages()) == FALSE) {install.packages("nycflights13")}
data = mtcars

Creating plot object

plot = ggplot(
  data, aes(
    x = wt,
    y = mpg
  )
)
plot

The graph page has been made, but no data has been plotted on the graph yet, we need to specify the data

plot + geom_point()

Adding a smoother

plot + geom_point() + geom_smooth()

## `geom_smooth()` using method = 'loess' and formula 'y ~ x'

The above smoother is by default method = “loess”, using a linear model - lm smoother

plot + geom_point() + geom_smooth(method = "lm", se = F)

## `geom_smooth()` using formula 'y ~ x'

Adding facets

plot + geom_point() + facet_grid(.~cyl) + geom_smooth(method = "lm",se = F)

## `geom_smooth()` using formula 'y ~ x'

Annotations

• Labels: xlab(), ylab(), ggtitle()
• Two standard themes
  • theme_gray() - default
  • theme_bw()

plot + geom_point(size = 4,color = "steelblue", alpha = 0.5)

plot + geom_point(aes(color = factor(cyl)),size = 4, alpha = 0.5)

Custom colors

colors <- c("4" = "#D9717D", "6" = "#4DB6D0", "8" = "#BECA55")
plot + geom_point(aes(color = factor(cyl)),size = 4, alpha = 0.5) + scale_color_manual(values = colors)

Adding labels

plot + geom_point(aes(color = factor(cyl)),size = 4, alpha = 0.5) + labs(title = "Weight - Mileage Distribution", x = "weight", y = "mileage per gallon")

Dealing with scaling and cartesian coordinates

To create a line plot to analyse outlier within data

testdata = data.frame(x = 1:100, y = rnorm(100))
testdata[50,2] = 100
plot(testdata$x,testdata$y, type = "l")

To only look at the regular data using the base plot function

plot(testdata$x,testdata$y, type = "l", ylim = c(-3,3))

In case of ggplot adding the ylim() function to the plot first subsets the data to only include points within limit and then plot the data

• The original data:

testplot = ggplot(
  testdata,
  aes(x = x,
      y = y)
)
testplot + geom_line()

• adding ylim()

testplot + geom_line() + ylim(-3,3)

We can clearly see the data at x = 50 is missing, to solve this we add coord_cartesian() instead of ylim() directly

testplot + geom_line() + coord_cartesian(ylim = c(-3,3))

Dealing with continous variables during plot

In mnay cases we’d be required to plot by conditioning continous variables, there are several ways to go about this, one method is to categorise the varaible into reasonable series of ranges, we use the cut() function for this purpose

data_weather = nycflights13::weather
data_weather$month = factor(data_weather$month)
cutpoints_1 = quantile(data_weather$humid, seq(0,1,length = 4), na.rm = T)
cutpoints_2 = quantile(data_weather$day, seq(0,1,length = 3), na.rm = T)
data_weather$humid_levels = cut(data_weather$humid, cutpoints_1)
data_weather$hour_levels = cut(data_weather$hour, cutpoints_2)
levels(data_weather$humid_levels)

## [1] "(12.7,51.8]" "(51.8,72.5]" "(72.5,100]"

levels(data_weather$hour_levels)

## [1] "(1,16]"  "(16,31]"

Cleaning up the variable

lev_humid = levels(data_weather$humid_levels)
lev_hour = levels(data_weather$hour_levels)
data_weather$humid_levels = factor(data_weather$humid_levels, levels = lev_humid, labels = c("low","med","high"))
data_weather$hour_levels = factor(data_weather$hour_levels, levels = lev_hour, labels = c("day","night"))
levels(data_weather$humid_levels)

## [1] "low"  "med"  "high"

levels(data_weather$hour_levels)

## [1] "day"   "night"

Plotting the data

ggplot(
  data_weather[complete.cases(data_weather[-11]),],
  aes(
    x = temp,
    y = dewp
  )
) + geom_point(
  alpha = 0.5
) + facet_wrap(
  hour_levels~humid_levels,
  nrow = 2, 
  ncol = 3
) + geom_smooth(
  method = "lm",
  se = F,
  col = "steelblue"
) + labs(
  title = "Weight - Mileage Distribution",
  x = "Weight",
  y = "Mileage"
)

## `geom_smooth()` using formula 'y ~ x'

Cleaning up data