• Based on "The Grammar of Graphics"
• The fundament for plotting in R nowadays
• Well documented
  • http://ggplot2.tidyverse.org
  • http://ggplot2.org
  • complete book (online version)
• Getting help: ggplot2 mailing list
• An increasing number of ggplot2 extensions developed by R users in the community

• Start with ggplot(data, mapping = aes()) where you supply a dataset and (default) aesthetic mapping
• Add a layer by calling a geom_function
• Then add on (not required as defaults supplied)
  
  • scales like xlim()
  • faceting like facet_wrap()
    
  • coordinate systems like coord_flip()
  • themes like theme_bw()
• Save a plot to disk with ggsave()

source of image (topright): older version of Data Visualization with ggplot cheat sheet (licensed under CC-BY-SA)

• Start with ggplot(data, mapping = aes()) where you supply a dataset and (default) aesthetic mapping
• Add a layer by calling a geom_function
• Then add on (not required as defaults supplied)
  
  • scales like xlim()
  • faceting like facet_wrap()
    
  • coordinate systems like coord_flip()
  • themes like theme_bw()
• Save a plot to disk with ggsave()

• Start with ggplot(data, mapping = aes()) where you supply a dataset and (default) aesthetic mapping
• Add a layer by calling a geom_function
• Then add on (not required as defaults supplied)
  
  • scales like xlim()
  • faceting like facet_wrap()
    
  • coordinate systems like coord_flip()
  • themes like theme_bw()
• Save a plot to disk with ggsave()

• Start with ggplot(data, mapping = aes()) where you supply a dataset and (default) aesthetic mapping
• Add a layer by calling a geom_function
• Then add on (not required as defaults supplied)
  
  • scales like xlim()
  • faceting like facet_wrap()
    
  • coordinate systems like coord_flip()
  • themes like theme_bw()
• Save a plot to disk with ggsave()

• Start with ggplot(data, mapping = aes()) where you supply a dataset and (default) aesthetic mapping
• Add a layer by calling a geom_function
• Then add on (not required as defaults supplied)
  
  • scales like xlim()
  • faceting like facet_wrap()
    
  • coordinate systems like coord_flip()
  • themes like theme_bw()
• Save a plot to disk with ggsave()

• Start with ggplot(data, mapping = aes()) where you supply a dataset and (default) aesthetic mapping
• Add a layer by calling a geom_function
• Then add on (not required as defaults supplied)
  
  • scales like xlim()
  • faceting like facet_wrap()
    
  • coordinate systems like coord_flip()
  • themes like theme_bw()
• Save a plot to disk with ggsave()
  

• Start with ggplot(data, mapping = aes()) where you supply a dataset and (default) aesthetic mapping
• Add a layer by calling a geom_function
• Then add on (not required as defaults supplied)
  
  • scales like xlim()
  • faceting like facet_wrap()
    
  • coordinate systems like coord_flip()
  • themes like theme_bw()
• Save a plot to disk with ggsave()
  

• Start with ggplot(data, mapping = aes()) where you supply a dataset and (default) aesthetic mapping
• Add a layer by calling a geom_function
• Then add on (not required as defaults supplied)
  
  • scales like xlim()
  • faceting like facet_wrap()
    
  • coordinate systems like coord_flip()
  • themes like theme_bw()
• Save a plot to disk with ggsave()
  

ggplot(iris, aes(x = Sepal.Length, 
    y = Petal.Length))

ggplot(iris, aes(x = Sepal.Length, 
    y = Petal.Length)) +
  geom_point()

ggplot(iris, aes(x = Sepal.Length, 
    y = Petal.Length)) +
  geom_point(aes(col = Species))

ggplot(iris, aes(x = Sepal.Length, 
    y = Petal.Length)) +
  geom_point(aes(col = Species)) +
  geom_smooth()

ggplot(iris, aes(x = Sepal.Length, 
    y = Petal.Length)) +
  geom_point(aes(col = Species)) +
  geom_smooth(aes(col = Species), 
    method = "lm")

ggplot(iris, aes(x = Sepal.Length, 
    y = Petal.Length, col = Species)) +
  geom_point() +
  geom_smooth(method = "lm")

ggplot(iris, aes(x = Sepal.Length, 
    y = Petal.Length, col = Species)) +
  geom_point() +
  geom_smooth(method = "lm") +
  scale_colour_brewer() 

ggplot(iris, aes(x = Sepal.Length, 
    y = Petal.Length, col = Species)) +
  geom_point() +
  geom_smooth(method = "lm") +
  scale_colour_brewer() +
  facet_wrap(~Species, nrow=3)

ggplot(iris, aes(x = Sepal.Length, 
    y = Petal.Length, col = Species)) +
  geom_point() +
  geom_smooth(method = "lm") +
  scale_colour_brewer() +
  facet_wrap(~Species, nrow=3) +
  coord_polar()

ggplot(iris, aes(x = Sepal.Length, 
    y = Petal.Length, col = Species)) +
  geom_point() +
  geom_smooth(method = "lm") +
  scale_colour_brewer() +
  facet_wrap(~Species, nrow=3) +
  coord_polar() +
  theme_dark()

BARPLOTS -
are used for categorical or discrete variables. Bars do not touch each other; there are no ‘in-between’ values.

HISTOGRAMS and DENSITY PLOTS (can be combined) are used for continuous variables and are often used to check whether variables are normally distributed. Bars touch each other in histograms.

• differences in average y values between groups and whether these might be significant,
• if group variances differ,
• whether individual groups are normally distributed
• identify outlier

Click here to read how

SCATTERPLOTS -
are the most basic plots for continuous variables. They are also the least interpreting plots as they show every observation in the 2-dimensional space.

Another useful feature is that they can be combined with other plotting elements: defining aesthetics for a 3rd variable (e.g. colours of points) or adding regression or smoothing lines to help visualise the relationship:

hydro_sub <- hydro %>%
 select(fmonth, station, date_time) %>%
  # (fmonth = month as factor)
 distinct()

ggplot(hydro_sub,aes(x=fmonth)) +
  geom_bar()

hydro_sub <- hydro %>%
 select(fmonth,station,lat,long) %>%
 distinct()

ggplot(hydro_sub, aes(x = long,  
    y = lat, col = fmonth)) +
  geom_point() 

p <- hydro %>% ggplot(aes(x = temp)) +
  geom_histogram()
p

ggplot(hydro, 
  aes(x = psal, 
    y = temp, 
    col = day)) +
  geom_point() +
  facet_wrap(
    ~fmonth, 
    nrow = 3)

hydro_sub <- hydro %>% 
  filter(station=="0403",fmonth==7,day==11) 
p_temp <- ggplot(hydro_sub, aes(y=pres)) +
  geom_point(aes(x = temp), col="red") +
  ylim(70, 0)
p_sal <- ggplot(hydro_sub, aes(y=pres)) +
  geom_point(aes(x = psal)) +
  ylim(70, 0)
p_oxy <- ggplot(hydro_sub, aes(y=pres)) +
  geom_point(aes(x = doxy), col="blue") +
  ylim(70, 0)
gridExtra::grid.arrange(grobs = list(
  p_temp, p_sal, p_oxy), nrow=1)

library(ggridges)
hydro %>% filter(pres < 10) %>%
  group_by(station, fmonth, day) %>%
  summarise(
    sst = mean(temp, na.rm = TRUE)) %>%
  group_by(station, fmonth) %>%
  summarise(
    sst = mean(sst, na.rm = TRUE)) %>%
  ungroup() %>%
  ggplot(aes(x = sst, y = fmonth, 
    fill = fmonth)) +
  geom_density_ridges(scale = 12, 
    rel_min_height = 0.005) +
  scale_fill_cyclical(
    values = c("blue", "green")) +
  theme_ridges()

library(ggmap)
b <- matrix(c(
  min(hydro$long), 
  max(hydro$long), 
  min(hydro$lat), 
  max(hydro$lat) ), byrow=T,nrow=2)
colnames(b) <- c("min","max")
rownames(b) <- c("x","y")
map_bs <- ggmap(get_map(location = b, 
  zoom = 5))
map_bs + geom_point(data=hydro, 
  aes(long,lat), size=0.5, 
  color="red")

baltic <- c(left = min(hydro$long), 
            bottom = min(hydro$lat),
            right = max(hydro$long), 
            top = max(hydro$lat))
map_stamen <- get_stamenmap(baltic, 
  zoom = 5, maptype = "toner-lite")
ggmap(map_stamen) + 
  geom_point(data=hydro, 
  aes(long,lat), size=0.5, 
  color="red") + theme_linedraw()

1. What happens if you make a scatterplot of station (x) vs temp (y)? Why is the plot not useful? What would be a better plot?
2. What happens if you make a boxplot of cruise (x) vs psal (y)? Why is this plot less suitable? What could be an alternative?