Basic R has multiple, separate functions, each used for creating a specific type of representation: boxplot, histogram, scatter plot etc. ggplot2 is an R package for creating elegant data visualization using the conceptual philosophy that views a plot as the assembly of different fundamental parts: \[Plot = Data + Aesthetics + Geometry\]

• Plot: the final graphics.
  
• Data: tabular data in tibble or a data frame.
  
• Aesthetics: Describe visual characteristics that represent data (position, size, color, shape, transparency, fill, scales.
  
• Geometry: defines the graphical representation: histogram, boxplot, scatter plot. Defines the type of geometric objects that represent data (points, lines, polygons.).

ggplot2 constructs graphs in layers using a grammar of graphics.

Each element is built as a layer based on a “grammar of graphics” all assembled into a final plot.

The “grammar” contains more definitions for graphics elements.

• coordinate system: e.g. Cartesian, polar, map projections.
  
• geoms: describe type of geometric objects that represent data (points, lines, polygons.
  
• aesthetics: describe visual characteristics that represent data (position, size, color, shape, transparency, fill.).
  
• scales: for each aesthetic: log scales, color scales, size scales, shape scales.
  
• stats : describe statistical transformations that typically summarize data: counts, means, medians, regression lines.
  
• facets: describe how data is split into subsets and displayed as multiple, separate small graphs.
  
• Theme: controls appearance of non-data elements.

In his 2017 presentation Hadley Wickham mentions thatggplot was created before Tidyverse and lacks the Tidyverse philosophy on the ideas of distinguishing and separating command (action) and query (computation) functions.(See references in tidyverse another R universe.)

However it is well integrated within the Tidyverse and can be placed at the end of a %>% pipeline as the last command.

There are many tutorials online to learn how to use ggplot.

See Appendix “ggplot2 Tutorials” for a table of just a few that seem useful based on the number of examples with ggplot code.

There are many more to be found with a simple web search.

One suggestion is The Complete ggplot2 Tutorial¹ split over multiple documents.

• A bar chart can be created by two types of geom: geom_bar() and geom_col().

• This simple knowledge can save you hours of frustration (see help with ?geom_bar().)

• Categorical variables are usually recognized automatically, but numerical and continuous variables have to be “made” into categories (or “levels”) by using as.factor() but in some cases as.character() might also work depending on the variable in question.

The dplyr chapter ended with the creation putting together a pipeline to create a summary data table.

The story will continue here as that chapter ended.

Splitting the pipeline above is most useful for using qplot or ggpolot.

Example 2 pipeline at midpoint before summarization, saved in object Mid.

It is the same code as above but stopped where the midpoint was suggested.

Mid <- Master4 %>% 
## select columns 
select(SEQN, RIAGENDR, RIDAGEYR, LBXMFOS, URXUCR, LBXTC ) %>% 
## fitler all rows to remove NAs 
drop_na() %>%  
## Creatinine adjustment 
mutate (RATIO = (LBXMFOS/URXUCR)*10^-4) %>% 
## categorize ages in  5 groups: 
##  Children: G0TO18, younger adults: G19TO35, 
## and older adults: G36TO65, seniors: G66TO79, 
## and 80 and older: G80. 
mutate(AGEGROUP = ifelse(RIDAGEYR %in% 0:18, "G0TO18", 
                ifelse(RIDAGEYR %in% 19:35, "G19TO35", 
                        ifelse(RIDAGEYR %in% 36:65, "G36TO65", 
                               ifelse(RIDAGEYR %in% 66:79,  
                                      "G66TO79", "G80")))))  

Below are some plot examples using Mid.

The addition of facet_grid splits the data “as a function of” (~) gender in RIAGENDR.

## Qplot (deprecated since versio. 3.4.0)
# qplot(AGEGROUP, data = Mid, geom="bar") 

# ggplot
ggplot(data = Mid, aes(x = AGEGROUP)) +
  geom_bar()

The geom_bar() function is used to create bar plots. The aes() function is used to map variables in the data to visual properties (aesthetics) of the plot. In this case, x = AGEGROUP is mapping the AGEGROUP variable to the x-axis. The + operator is used to add layers to the plot.

It would be useful to visualize based on gender, accomplished with the facet_grid() function.

## Qplot 
# qplot(AGEGROUP, data = Mid, geom="bar") +  
#   facet_grid(~RIAGENDR) 

ggplot(data = Mid, aes(x = AGEGROUP)) +
  geom_bar() +
  facet_grid(~RIAGENDR)

To add color we need to use geom_bar instead of geom = "bar" so that we can add an aesthetics (aes) request to color, as a factor of the values in RIAGENDR.

ggplot(data = Mid, aes(x = AGEGROUP)) +
  geom_bar(aes(fill = as.factor(RIAGENDR))) + facet_grid(~RIAGENDR)

With facet_grid() the age distribution by gender on two separate graphs.

A similar plot but with stacked bars can be achieved with ggplot.

We can avoid using as.factor that is necessary since RIAGENDR is coded as a number that ggplot considers a numerical (perhaps continuous) rather than a categorical variable.

We could avoid this problem by “recoding” the values of 1 and 2 to words such as male and female or Men and Women on a short pipeline before the plot is done.

(Review recode() in section “recoding string replacement.)

Mid %>%   
  mutate(RIAGENDR =  
           recode(RIAGENDR,  
                  `1` =  "Men",  
                  `2` = "Women")) %>%  
ggplot(aes(x = AGEGROUP)) +  
  geom_bar(aes(fill = RIAGENDR)) 

(Plot on next slide.)

We now also have a better description, avoiding 1 and 2 as well as as.factor in the legend.

A final touch could be to rename the column RIAGENDR to simply Gender and AGEGROUP to Age group by using the rename() function (section @ref(recodingstringreplacement).).

We can saved this in Mid2.

Note the need of quote for Age group to take care of the blank space.

Mid2 <- Mid %>% 
    mutate(RIAGENDR =  
           recode(RIAGENDR,  
                  `1` =  "Men",  
                  `2` = "Women")) %>%  
    rename(Gender  = RIAGENDR, "Age group" =  AGEGROUP) 

To have the bars side by side for each age group the additional position = option is introduced with option "dodge" (bars touch) or "dodge2" (white space between bars.)

## 
Mid2 %>% ggplot(aes(x = `Age group`)) +  
  geom_bar(aes(fill = Gender), position = "dodge2") 

Side by side bar of gender count by age group is possible with the dodge or dodge2 options.

It is possible to combine options:

 Mid2 %>%  
  ggplot(aes(x = `Age group`)) +  
  geom_bar(aes(fill = Gender), position = "dodge2") +  facet_wrap(~ `Age group`) 

Side by side in each facet.

Example derived from info at Plotting with ggplot: bar plots with error bars (See also Appendix “ggplot2 Tutorials.”)

We need to use ungroup() as data were grouped when creating Example2. (Section “by gender and age.”)

Example2 %>% 
## ungroup to allow changes for mutate and rename 
    ungroup() %>% 
      mutate(RIAGENDR =  
           recode(RIAGENDR,  
                  `1` =  "Men",  
                  `2` = "Women")) %>%  
  rename(Gender  = RIAGENDR) %>% 
## start ggplot commands 
  ggplot(aes(AGEGROUP, MeanTChol)) +  
  geom_col(aes(fill = Gender)) +   
  geom_errorbar(aes(ymin = MeanTChol - sdTChol,  
                    ymax = MeanTChol + sdTChol),  
                width=0.3) +  
  facet_wrap(~Gender) + 
  labs(y="Mean Total Cholesterol ± s.d. (mg/dL)", x = "Age by group")