Counts

How Many of this and that?

Arvind V.

2024-06-23

You’re Nothing but a Pack of Cards!!

You’re Nothing but a Pack of Cards!!

“No matter what happens in life, be good to people. Being good to people is a wonderful legacy to leave behind.”

— Taylor Swift

Setting up R Packages

library(tidyverse) # Sine qua non
library(mosaic) # Out all-in-one package
library(ggformula) # Graphing package
library(skimr) # Looking at Data
library(janitor) # Clean the data
library(naniar) # Handle missing data
library(visdat) # Visualise missing data
library(tinytable) # Printing Static Tables for our data
library(DT) # Interactive Tables for our data
library(crosstable) # Multiple variable summaries
library(marquee) # For Annotations with Fonts
library(ggrepel) # Repel overlapping text labels in ggplot2

Plot Fonts and Theme

Code 
library(systemfonts)
library(showtext)
## Clean the slate
systemfonts::clear_local_fonts()
systemfonts::clear_registry()
##
showtext_opts(dpi = 96) # set DPI for showtext
sysfonts::font_add(
  family = "Alegreya",
  regular = "../../../../../../fonts/Alegreya-Regular.ttf",
  bold = "../../../../../../fonts/Alegreya-Bold.ttf",
  italic = "../../../../../../fonts/Alegreya-Italic.ttf",
  bolditalic = "../../../../../../fonts/Alegreya-BoldItalic.ttf"
)

sysfonts::font_add(
  family = "Roboto Condensed",
  regular = "../../../../../../fonts/RobotoCondensed-Regular.ttf",
  bold = "../../../../../../fonts/RobotoCondensed-Bold.ttf",
  italic = "../../../../../../fonts/RobotoCondensed-Italic.ttf",
  bolditalic = "../../../../../../fonts/RobotoCondensed-BoldItalic.ttf"
)
showtext_auto(enable = TRUE) # enable showtext
##
theme_custom <- function() {
  theme_bw(base_size = 10) +

    # theme(panel.widths = unit(11, "cm"),
    #       panel.heights = unit(6.79, "cm")) + # Golden Ratio

    theme(
      plot.margin = margin_auto(t = 1, r = 2, b = 1, l = 1, unit = "cm"),
      plot.background = element_rect(
        fill = "bisque",
        colour = "black",
        linewidth = 1
      )
    ) +

    theme_sub_axis(
      title = element_text(
        family = "Roboto Condensed",
        size = 10
      ),
      text = element_text(
        family = "Roboto Condensed",
        size = 8
      )
    ) +

    theme_sub_legend(
      text = element_text(
        family = "Roboto Condensed",
        size = 6
      ),
      title = element_text(
        family = "Alegreya",
        size = 8
      )
    ) +

    theme_sub_plot(
      title = element_text(
        family = "Alegreya",
        size = 14, face = "bold"
      ),
      title.position = "plot",
      subtitle = element_text(
        family = "Alegreya",
        size = 10
      ),
      caption = element_text(
        family = "Alegreya",
        size = 6
      ),
      caption.position = "plot"
    )
}

## Use available fonts in ggplot text geoms too!
ggplot2::update_geom_defaults(geom = "text", new = list(
  family = "Roboto Condensed",
  face = "plain",
  size = 3.5,
  color = "#2b2b2b"
))
ggplot2::update_geom_defaults(geom = "label", new = list(
  family = "Roboto Condensed",
  face = "plain",
  size = 3.5,
  color = "#2b2b2b"
))

ggplot2::update_geom_defaults(geom = "marquee", new = list(
  family = "Roboto Condensed",
  face = "plain",
  size = 3.5,
  color = "#2b2b2b"
))
ggplot2::update_geom_defaults(geom = "text_repel", new = list(
  family = "Roboto Condensed",
  face = "plain",
  size = 3.5,
  color = "#2b2b2b"
))
ggplot2::update_geom_defaults(geom = "label_repel", new = list(
  family = "Roboto Condensed",
  face = "plain",
  size = 3.5,
  color = "#2b2b2b"
))

## Set the theme
ggplot2::theme_set(new = theme_custom())

## tinytable options
options("tinytable_tt_digits" = 2)
options("tinytable_format_num_fmt" = "significant_cell")
options(tinytable_html_mathjax = TRUE)


## Set defaults for flextable
flextable::set_flextable_defaults(font.family = "Roboto Condensed")

What graphs will we see today?

Variable #1 Variable #2 Chart Names Chart Shape
Qual None Bar Chart

What kind of Data Variables will we choose?

No Pronoun Answer Variable/Scale Example What Operations?
3 How, What Kind, What Sort A Manner / Method, Type or Attribute from a list, with list items in some " order" ( e.g. good, better, improved, best..) Qualitative/Ordinal Socioeconomic status (Low income, Middle income, High income),Education level (HighSchool, BS, MS, PhD),Satisfaction rating(Very much Dislike, Dislike, Neutral, Like, Very Much Like) Median,Percentile

Inspiration: Column Chart

Figure 1: Capital Cities

How much does the (financial) capital of a country contribute to its GDP? Which would be India’s city? What would be the reduction in percentage? And these Germans are crazy. (Toc, toc, toc, toc!)

Note how the axis variable that defines the bar locations is a …Qual variable!

ggformula and mosaic API

  • Recall the API: the programming interface to each of mosaic, ggformula, and ggplot.
  • As stated earlier, mosaic and ggformula have a very similar, and intuitive, interface.

Tip

Note the standard method for all commands from the {mosaic} and {ggformula} packages: goal( y ~ x | z, data = _____)

With {mosaic}, one can create a statistical correlation test between two variables as: cor_test(y ~ x, data = ______ )

With {ggformula}, one can create any graph/chart using: gf_***(y ~ x | z, data = _____) - In practice, we often use: dataframe %>% gf_***(y ~ x | z) which has cool benefits such as “autocompletion” of variable names. - The ” *** ” indicates what kind of graph you desire: histogram, bar, scatter, density; - The “——-” is the name of your dataset that you want to plot with.

ggplot API

ggplot command template

ggplot(data = ---, mapping = aes(x = ---, y = ---)) + geom_----()

  • ” —- ” is meant to imply text you supply. e.g. function names, data frame names, variable names.
  • It is helpful to see the argument mapping, above.
  • In practice, rather than typing the formal arguments…
  • ggplot code is typically shorthanded to this:

dataframe %>% ggplot(aes(xvar, yvar)) + geom_----() - Note the change from %>% to + when adding a geom. Sigh.

Bar Charts and Histograms

Bar Charts and Histograms: Similar but Different

  • Bar Charts show counts of observations with respect to a Qualitative variable.
  • For instance, a shop inventory with shirt-sizes.
  • Each bar has a height proportional to the count per shirt-size, in this example.
  • Although Histograms may look similar to Bar Charts, the two are different.
  • First, histograms show continuous Quant data.
  • By contrast, bar charts show categorical data, such as shirt-sizes, or apples, bananas, carrots, etc.
  • Visually speaking, histograms do not usually show spaces between bars because these are continuous values,
  • While column charts must show spaces to separate each category.

How do Bar Chart(s) Work?

  • Bar are used to show “counts” and “tallies” with respect to Qual variables: they answer the question How Many?.
  • For instance, in a survey, how many people vs Gender?
  • In a Target Audience survey on Weekly Consumption, how many low, medium, or high expenditure people?
  • Each Qual variable potentially has many levels as we saw in the Nature of Data.
  • In Weekly Consumption, low, medium and high were levels for the Qual variable Expenditure.
  • Bar charts perform internal counts for each level of the Qual variable under consideration.
  • The Bar Plot is then a set of disjoint bars representing these counts; see the icon above, and then that for histograms!!
  • The X-axis is the set of levels in the Qual variable, and the Y-axis represents the counts for each level.

Case Study-1: Chicago Taxi Rides dataset

Read Data

We will first look at at a dataset that speaks about taxi rides in Chicago in the year 2022. This is available on Vincent Arel-Bundock’s superb repository of datasets. Let us read into R directly from the website.

Code 
taxi <- read_csv("https://vincentarelbundock.github.io/Rdatasets/csv/modeldata/taxi.csv")

taxi_modified <- taxi %>%
  naniar::replace_with_na_all(condition = ~ .x %in% common_na_strings) %>%
  naniar::replace_with_na_all(condition = ~ .x %in% common_na_numbers) %>%
  janitor::clean_names(case = "snake") %>%
  janitor::remove_empty()

taxi_modified

The data has automatically been read into the webr session, so you can continue on to the next code chunk!

Examine the Data

As per our Workflow, we will look at the data using all the three methods we have seen.

dplyr::glimpse(taxi)
Rows: 10,000
Columns: 8
$ rownames <dbl> 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18…
$ tip      <chr> "yes", "yes", "yes", "yes", "yes", "yes", "yes", "yes", "yes"…
$ distance <dbl> 17.19, 0.88, 18.11, 20.70, 12.23, 0.94, 17.47, 17.67, 1.85, 1…
$ company  <chr> "Chicago Independents", "City Service", "other", "Chicago Ind…
$ local    <chr> "no", "yes", "no", "no", "no", "yes", "no", "no", "no", "no",…
$ dow      <chr> "Thu", "Thu", "Mon", "Mon", "Sun", "Sat", "Fri", "Sun", "Fri"…
$ month    <chr> "Feb", "Mar", "Feb", "Apr", "Mar", "Apr", "Mar", "Jan", "Apr"…
$ hour     <dbl> 16, 8, 18, 8, 21, 23, 12, 6, 12, 14, 18, 11, 12, 19, 17, 13, …
skimr::skim(taxi_modified) %>% as_tibble()
taxi_inspect <- mosaic::inspect(taxi_modified)
taxi_inspect$categorical
taxi_inspect$quantitative

Data Dictionary

Quantitative Data

  • distance: Continuous Quant variable, the distance of the trip in miles.

Qualitative Data

  • tip: Yes/No type Qual variable, whether a tip was given or not.
  • company: 7 levels, the cab company that was used for the ride.
  • local: 2 levels, whether the trip was local or not.
  • hour : 24 levels, the hour of the day when the trip started.
  • dow: 7 levels, the day of the week.
  • month: 12 levels, the month of the year.

Business Insights on Examining the taxi dataset

  • This is a large dataset (10K rows), 8 columns/variables.
  • There are several Qualitative variables: tip(2), company(7) and local(2), dow(7), and month(12). These have levels as shown in the parenthesis.
  • Note that hour despite being a discrete/numerical variable, it can be treated as a Categorical variable too.
  • distance is Quantitative.
  • There are no missing values for any variable, all are complete with 10K entries.

Data Munging

We will convert the tip, company, dow, local, hour, and month variables into factors beforehand.

Code 
## Convert `dow`, `local`, `month`, and `hour` into ordered factors
taxi_modified <- taxi_modified %>%
  dplyr::mutate(
    ## Variable "tip"
    tip = base::factor(tip,
      levels = c("yes", "no"),
      labels = c("yes", "no"),
      ordered = TRUE
    ),

    ## Variable "company"
    company = base::factor(company), # Any order is OK.

    ## Variable "dow"
    dow = base::factor(dow,
      levels = c("Mon", "Tue", "Wed", "Thu", "Fri", "Sat", "Sun"),
      labels = c("Mon", "Tue", "Wed", "Thu", "Fri", "Sat", "Sun"),
      ordered = TRUE
    ),

    ## Variable "local"
    local = base::factor(local,
      levels = c("yes", "no"),
      labels = c("yes", "no"),
      ordered = TRUE
    ),

    ## Variable "month"
    month = base::factor(month,
      levels = c("Jan", "Feb", "Mar", "Apr"),
      labels = c("Jan", "Feb", "Mar", "Apr"),
      ordered = TRUE
    ),

    ## Variable "hour"
    hour = base::factor(hour,
      levels = c(0:23), labels = c(0:23),
      ordered = TRUE
    )
  ) %>%
  dplyr::relocate(where(is.factor), .after = rownames) # Move all factors to the left

taxi_modified %>% glimpse()
Rows: 10,000
Columns: 8
$ rownames <dbl> 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18…
$ tip      <ord> yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, y…
$ company  <fct> Chicago Independents, City Service, other, Chicago Independen…
$ local    <ord> no, yes, no, no, no, yes, no, no, no, no, no, no, no, yes, no…
$ dow      <ord> Thu, Thu, Mon, Mon, Sun, Sat, Fri, Sun, Fri, Tue, Tue, Sun, W…
$ month    <ord> Feb, Mar, Feb, Apr, Mar, Apr, Mar, Jan, Apr, Mar, Mar, Apr, A…
$ hour     <ord> 16, 8, 18, 8, 21, 23, 12, 6, 12, 14, 18, 11, 12, 19, 17, 13, …
$ distance <dbl> 17.19, 0.88, 18.11, 20.70, 12.23, 0.94, 17.47, 17.67, 1.85, 1…

Looks clean and good.

Hypothesis and Research Questions

  • It is a good practice in Exploratory Data Analysis to surmise the Experiment that lead to the gathering of this dataset
  • And what the target variable might be.
  • This is the variable that is to be explained, or predicted, or modelled.
  • The other variables are explanatory variables, or predictor variables.
  • The target variable for an experiment that resulted in this data might be the tip variable
  • Since that looks like a response, or an outcome.
  • It is a binary i.e. Yes/No type Qual variable.
  • We will concentrate on the tip target variable to ask questions about the data, and then plot the answers to these questions.

Research Questions

  1. Do more people tip than not?
  2. Does a tip depend upon whether the trip is local or not?
  3. Do some cab company-ies get more tips than others?
  4. And does a tip depend upon the distance, hour of day, and dow and month?

Try and think of more Questions!

Plotting Barcharts

Let’s plot some bar graphs: recall that for bar charts, we need to choose Qual variables to count with! In each case, we will state a Hypothesis/Question and try to answer it with a chart.

Question-1: Do more people tip than not?

ggformula-1

Code 
ggplot2::theme_set(new = theme_custom())

gf_bar(~tip, data = taxi_modified) %>% #<<
  gf_labs(title = "Plot 1A: Counts of Tips")

ggplot-1

Code 
ggplot2::theme_set(new = theme_custom())

ggplot(taxi_modified) +
  geom_bar(aes(x = tip)) +
  labs(title = "Plot 1A: Counts of Tips")

Business Insights-1

  • Far more people do tip than not. Which is nice.
  • (Future) The counts of tip are very imbalanced and if we are to setup a model for that (logistic regression) we would need to very carefully subset the data for training and testing our model.

Question-2: Does the tip depend upon whether the trip is local or not?

ggformula-2

ggplot2::theme_set(new = theme_custom())

taxi_modified %>%
  gf_bar(~local,
    fill = ~tip,
    position = "dodge"
  ) %>%
  gf_labs(title = "Plot 2A: Dodged Bar Chart") %>%
  gf_refine(scale_fill_brewer(palette = "Set1"))

ggplot2::theme_set(new = theme_custom())

taxi_modified %>%
  gf_bar(~local,
    fill = ~tip,
    position = "stack"
  ) %>%
  gf_labs(
    title = "Plot 2B: Stacked Bar Chart",
    subtitle = "Can we spot per group differences in proportions??"
  ) %>%
  gf_refine(scale_fill_brewer(palette = "Set1"))

ggplot2::theme_set(new = theme_custom())

## Showing "per capita" percentages
taxi_modified %>%
  gf_bar(~local,
    fill = ~tip,
    position = "fill"
  ) %>%
  gf_labs(
    title = "Plot 2C: Filled Bar Chart",
    subtitle = "Shows Per group differences in Proportions!"
  ) %>%
  gf_refine(scale_fill_brewer(palette = "Set1"))

ggplot2::theme_set(new = theme_custom())

## Showing "per capita" percentages
## Better labelling of Y-axis
taxi_modified %>%
  gf_props(~local,
    fill = ~tip,
    position = "fill"
  ) %>%
  gf_labs(
    title = "Plot 2D: Filled Bar Chart",
    subtitle = "Shows Per group differences in Proportions!"
  ) %>%
  gf_refine(scale_fill_brewer(palette = "Set1"))

ggplot-2

Code 
ggplot2::theme_set(new = theme_custom())

taxi_modified %>%
  ggplot() +
  geom_bar(aes(x = local, fill = tip), position = "dodge") +
  labs(title = "Plot 2A:Dodged Bar Chart") +
  scale_fill_brewer(palette = "Set1")
##
taxi_modified %>%
  ggplot() +
  geom_bar(aes(x = local, fill = tip), position = "stack") +
  labs(
    title = "Plot 2B: Stacked Bar Chart",
    subtitle = "Can we spot per group differences in proportions??"
  ) +
  scale_fill_brewer(palette = "Set1")
## Showing "per capita" percentages
taxi_modified %>%
  ggplot() +
  geom_bar(aes(x = local, fill = tip), position = "fill") +
  labs(title = "Plot 2C: Filled Bar Chart", subtitle = "Shows Per group differences in Proportions!") +
  scale_fill_brewer(palette = "Set1")
## Showing "per capita" percentages
## Better labelling of Y-axis
taxi_modified %>%
  ggplot() +
  geom_bar(aes(x = local, fill = tip), position = "fill") +
  labs(
    title = "Plot 2D: Filled Bar Chart",
    subtitle = "Shows Per group differences in Proportions!",
    y = "Proportion"
  ) +
  scale_fill_brewer(palette = "Set1")

Business Insights-2

  • Counting the frequency of tip by local gives us grouped counts, but we cannot tell the percentage per group (local or not) of those who tip and those who do not.
  • We need per-group percentages because the number of local trips are not balanced
  • With {ggformula}, we tried bar charts with position = stack, but finally it is the position = fill that works best.
  • We see that the percentage of tippers is somewhat higher with people who make non-local trips. Not surprising.

Question-3: Do some cab company-ies get more tips than others?

ggformula-3

ggplot2::theme_set(new = theme_custom())

taxi_modified %>%
  gf_bar(~company, fill = ~tip, position = "dodge") %>%
  gf_labs(title = "Plot 3A: Dodged Bar Chart") %>%
  gf_theme(theme(axis.text.x = element_text(
    size = 6,
    angle = 45, hjust = 0.5
  ))) %>%
  gf_refine(scale_fill_brewer(palette = "Set1"))

ggplot2::theme_set(new = theme_custom())

taxi_modified %>%
  gf_bar(~company, fill = ~tip, position = "stack") %>%
  gf_labs(
    title = "Plot 3B: Stacked Bar Chart",
    subtitle = "Can we spot per group differences in proportions??"
  ) %>%
  gf_theme(theme(axis.text.x = element_text(size = 6, angle = 45, hjust = 1))) %>%
  gf_refine(scale_fill_brewer(palette = "Set1"))

ggplot2::theme_set(new = theme_custom())

## Showing "per capita" percentages
taxi_modified %>%
  gf_percents(~company, fill = ~tip, position = "fill") %>%
  gf_labs(
    title = "Plot 3C: Filled Bar Chart",
    subtitle = "Shows Per group differences in Proportions!"
  ) %>%
  gf_theme(theme(axis.text.x = element_text(size = 6, angle = 45, hjust = 1))) %>%
  gf_refine(scale_fill_brewer(palette = "Set1"))

ggplot2::theme_set(new = theme_custom())

## Showing "per capita" percentages
## Better labelling of Y-axis
taxi_modified %>%
  gf_props(~company, fill = ~tip, position = "fill") %>%
  gf_labs(
    title = "Plot 3D: Filled Bar Chart",
    subtitle = "Shows Per group differences in Proportions!"
  ) %>%
  gf_theme(theme(axis.text.x = element_text(size = 6, angle = 45, hjust = 1))) %>%
  gf_refine(scale_fill_brewer(palette = "Set1"))

ggplot-3

Code 
ggplot2::theme_set(new = theme_custom())

taxi_modified %>%
  ggplot() +
  geom_bar(aes(x = company, fill = tip), position = "dodge") +
  labs(title = "Plot 3A: Dodged Bar Chart") +
  theme(theme(axis.text.x = element_text(size = 6, angle = 45, hjust = 1))) +
  scale_fill_brewer(palette = "Set1")
##
taxi_modified %>%
  ggplot() +
  geom_bar(aes(x = company, fill = tip), position = "stack") +
  labs(
    title = "Plot 3B: Stacked Bar Chart",
    subtitle = "Can we spot per group differences in proportions??"
  ) +
  theme(theme(axis.text.x = element_text(size = 6, angle = 45, hjust = 1))) +
  scale_fill_brewer(palette = "Set1")
## Showing "per capita" percentages
taxi_modified %>%
  ggplot() +
  geom_bar(aes(x = company, fill = tip), position = "fill") +
  labs(
    title = "Plot 3C: Filled Bar Chart",
    subtitle = "Shows Per group differences in Proportions!"
  ) +
  theme(theme(axis.text.x = element_text(size = 6, angle = 45, hjust = 1))) +
  scale_fill_brewer(palette = "Set1")
## Showing "per capita" percentages
## Better labelling of Y-axis
taxi_modified %>%
  ggplot() +
  geom_bar(aes(x = company, fill = tip), position = "fill") +
  labs(
    title = "Plot 3D: Filled Bar Chart",
    subtitle = "Shows Per group differences in Proportions!",
    y = "Proportions"
  ) +
  theme(theme(axis.text.x = element_text(size = 6, angle = 45, hjust = 1))) +
  scale_fill_brewer(palette = "Set1")

Business Insights-3

  • Using stack-ed, dodge-ed, and fill-ed in {ggformula} in bar plots gives us different ways of looking at the sets of counts;
  • fill: gives us a per-group proportion of another Qual variable for a chosen Qual variable. This chart view is useful in Inference for Proportions;
  • Most cab company-ies have similar usage, if you neglect the other category of company;
  • Does seem that of all the company-ies, tips are not so good for the Flash Cab company. A driver issue? Or are the cars too old? Or don’t they offer service everywhere?

Question-4: Does a tip depend upon the distance, hour of day, and dow and month?

ggformula-4

ggplot2::theme_set(new = theme_custom())

gf_bar(~hour, fill = ~tip, data = taxi_modified) %>%
  gf_labs(title = "Plot 4A: Counts of Tips by Hour") %>%
  gf_refine(scale_fill_brewer(palette = "Set1"))

ggplot2::theme_set(new = theme_custom())

gf_bar(~dow, fill = ~tip, data = taxi_modified) %>%
  gf_labs(title = "Plot 4B: Counts of Tips by Day of Week") %>%
  gf_refine(scale_fill_brewer(palette = "Set1"))

ggplot2::theme_set(new = theme_custom())

gf_bar(~month, fill = ~tip, data = taxi_modified) %>%
  gf_labs(title = "Plot 4C: Counts of Tips by Month") %>%
  gf_refine(scale_fill_brewer(palette = "Set1"))

ggplot2::theme_set(new = theme_custom())

gf_bar(~ month | dow, fill = ~tip, data = taxi_modified) %>%
  gf_labs(title = "Plot 4D: Counts of Tips by Day of Week and Month") %>%
  gf_refine(scale_fill_brewer(palette = "Set1"))

ggplot2::theme_set(new = theme_custom())

## This may be too busy a graph...
gf_bar(~ dow | hour, fill = ~tip, data = taxi_modified) %>%
  gf_labs(
    title = "Plot 4E: Counts of Tips by Hour and Day of Week",
    subtitle = "Is this plot arrangement easy to grasp?"
  ) %>%
  gf_refine(scale_fill_brewer(palette = "Set1"))

ggplot2::theme_set(new = theme_custom())

## This is better!
gf_bar(~ hour | dow, fill = ~tip, data = taxi_modified) %>%
  gf_labs(
    title = "Plot 4F: Counts of Tips by Hour and Day of Week",
    subtitle = "Facetted by Day of Week"
  ) %>%
  gf_refine(scale_fill_brewer(palette = "Set1"))

ggplot-4

Code 
ggplot2::theme_set(new = theme_custom())

gf_bar(~hour, fill = ~tip, data = taxi_modified) %>%
  gf_labs(title = "Plot 4A: Counts of Tips by Hour") %>%
  gf_refine(scale_fill_brewer(palette = "Set1"))
##
ggplot(taxi_modified) +
  geom_bar(aes(x = dow, fill = tip)) +
  labs(title = "Plot 4B: Counts of Tips by Day of Week") +
  scale_fill_brewer(palette = "Set1")
##
ggplot(taxi_modified) +
  geom_bar(aes(x = month, fill = tip)) +
  labs(title = "Plot 4C: Counts of Tips by Month") +
  scale_fill_brewer(palette = "Set1")
##
ggplot(taxi_modified) +
  geom_bar(aes(x = month, fill = tip)) +
  facet_wrap(~dow) +
  labs(title = "Plot 4D: Counts of Tips by Day of Week and Month") +
  scale_fill_brewer(palette = "Set1")
##
ggplot(taxi_modified) +
  geom_bar(aes(x = dow, fill = tip)) +
  facet_wrap(~hour) +
  labs(
    title = "Plot 4E: Counts of Tips by Hour and Day of Week",
    subtitle = "Is this plot arrangement easy to grasp?"
  ) +
  scale_fill_brewer(palette = "Set1")
##
ggplot(taxi_modified) +
  geom_bar(aes(x = hour, fill = tip)) +
  facet_wrap(~dow) +
  labs(
    title = "Plot 4F: Counts of Tips by Hour and Day of Week",
    subtitle = "Swapped the Facets"
  ) +
  scale_fill_brewer(palette = "Set1")

Business Insights-4

  • Note: We were using fill = ~ tip here! Why is that a good idea?
  • tips vs hour: There are always more people who tip than those who do not. Of course there are fewer trips during the early morning hours and the late night hours, based on the very small bar-pairs we see at those times
  • tips vs dow: Except for Sunday, the tip count patterns (Yes/No) look similar across all days.
  • tips vs month: We have data for 4 months only. Again, the tip count patterns (Yes/No) look similar across all months. Perhaps slightly fewer trips in Jan, when it is cold in Chicago and people may not go out much.
  • tips vs dow vs month: Very similar counts for tips(Yes/No) across day-of-week and month.

Bar Plot Extras

gf-bar and gf-col

  • Note also that gf_bar/geom_bar takes only ONE variable (for the x-axis)
  • Whereas gf_col/geom_col needs both X and Y variables since it simply plots columns.
  • Both are useful!
  • We have already seen gf_props in our two case studies above.
  • Also check out gf_percents !

Proportions and Percentages

ggplot2::theme_set(new = theme_custom())

gf_props(~substance,
  data = mosaicData::HELPrct, fill = ~sex,
  position = "dodge"
) %>%
  gf_labs(
    title = "Plotting Proportions using gf_props",
    subtitle = "Option = dodge"
  ) %>%
  gf_refine(scale_fill_brewer(palette = "Set1"))

ggplot2::theme_set(new = theme_custom())

gf_props(~substance,
  data = mosaicData::HELPrct, fill = ~sex,
  position = "fill"
) %>%
  gf_labs(
    title = "Plotting Proportions using gf_props",
    subtitle = "Option = fill"
  ) %>%
  gf_refine(scale_fill_brewer(palette = "Set1"))

ggplot2::theme_set(new = theme_custom())

gf_percents(~substance,
  data = mosaicData::HELPrct, fill = ~sex,
  position = "dodge"
) %>%
  gf_refine(
    scale_y_continuous(
      labels = scales::label_percent(scale = 1)
    )
  ) %>%
  gf_labs(title = "Plotting Percentages using gf_percents") %>%
  gf_refine(scale_fill_brewer(palette = "Set1"))

Are the Differences in Proportion Significant?

  • When we see situations such as this, where data has one or more Qual variables that are binary(Yes/No)..
  • We are always interested in whether these proportions of Yes/No are really different
  • Or if we are just seeing the result of random chance.
  • This is usually mechanized by a Stat Test called a Single Proportion Test
  • Or, when we have more than one, a Multiple Proportion Test.

Your Turn

  1. Click on the Dataset Icon, and unzip that archive. Try to make Bar plots with each of them, using one or more Qual variables. Datasets
  2. A dataset from calmcode.io https://calmcode.io/datasets.html
  3. AiRbnb Price Data on the French Riviera.
  4. Apartment price vs ground living area.
  5. Fertility: This rather large and interesting Fertility related dataset from https://vincentarelbundock.github.io/Rdatasets/csv/AER/Fertility.csv
  6. Songs by Kishore Kumar: https://sunilslists.com/hindi-songs/luminaries-hindi-songs/kishore-kumar-songs-all

glimpse / skim / inspect the dataset in each case, state that Data Dictionary, and develop a set of Questions that can be answered by appropriate stat measures, or by using a chart to show the distribution.

Wait, But Why?

  • Always count your chickens count your data before you model or infer!
  • Counts first give you an absolute sense of how much data you have.
  • Counts by different Qual variables give you a sense of the combinations you have in your data: \((Male/Female) * (Income-Status) * (Old/Young) * (Urban/Rural)\) (Say 2 * 3 * 2 * 2 = 24 combinations of data)
  • Counts then give an idea whether your data is lop-sided: do you have too many observations of one category(level) and too few of another category(level) in a given Qual variable?
  • Balance is important in order to draw decent inferences
  • And for ML algorithms, to train them properly.

Counts from Literature

Figure 2: Zipf’s Law

Zipf’s Law

  • Since the X-axis in bar charts is Qualitative (the bars don’t touch, remember!) it is possible to sort the bars at will, based on the levels within the Qualitative variables. See the approx Zipf’s Law distribution for the English alphabet above
  • In Figure 2, the letters of the alphabet are “levels” within a Qualitative variable
  • these levels have been sorted based on the frequency or count!
  • This is what Sherlock Holmes might have done,
  • Or the method how they cracked the code to the treasure in this story.

Conclusion

  • Qualitative data variables can be plotted as counts, using Bar Charts
  • gf_col and gf_bar provide Bar charts; gf_bar performs counts internally, whereas gf_col requires pre-counted data.
  • gf_props and gf_percents provide Bar charts of proportions and percentages, respectively
  • position = "dodge" gives side-by-side bars for each level of a Qual variable
  • position = "stack" gives stacked bars for each level of a Qual variable
  • position = "fill" gives stacked bars scaled to 100% height, showing per-group proportions for each level of a Qual variable
  • facet_wrap(~ var) or facet_grid(var1 ~ var2) allows us to create multiple plots based on one or two other Qual variables

AI Generated Summary and Podcast

  • This text excerpt focuses on bar charts and histograms as visualization tools for qualitative and quantitative data, respectively. - It walks the reader through the creation of bar charts using the R programming language, illustrating the concept through a case study using the Chicago taxi rides dataset.
  • The author explores various scenarios and questions related to taxi tipping, such as the frequency of tips and their dependence on trip locality, company, hour of the day, and day of the week.
  • Finally, the excerpt highlights the importance of understanding data counts before undertaking data modeling or inference, emphasizing the role of bar charts in revealing data distribution and potential imbalances.

References

  1. Daniel Kaplan and Randall Pruim. ggformula: Formula Interface for ggplot2 (full version). https://www.mosaic-web.org/ggformula/articles/pkgdown/ggformula-long.html
  2. Winston Chang (2024). R Graphics Cookbook. https://r-graphics.org

R Package Citations

Package Version Citation
ggformula 1.0.0 Kaplan and Pruim (2025)
mosaic 1.9.2 Pruim, Kaplan, and Horton (2017)
tidyplots 0.3.1 Engler (2025)
tidyverse 2.0.0 Wickham et al. (2019)
tinyplot 0.5.0 McDermott, Arel-Bundock, and Zeileis (2025)
Engler, Jan Broder. 2025. “Tidyplots Empowers Life Scientists with Easy Code-Based Data Visualization.” iMeta, e70018. https://doi.org/10.1002/imt2.70018.
Kaplan, Daniel, and Randall Pruim. 2025. ggformula: Formula Interface to the Grammar of Graphics. https://doi.org/10.32614/CRAN.package.ggformula.
McDermott, Grant, Vincent Arel-Bundock, and Achim Zeileis. 2025. tinyplot: Lightweight Extension of the Base r Graphics System. https://doi.org/10.32614/CRAN.package.tinyplot.
Pruim, Randall, Daniel T Kaplan, and Nicholas J Horton. 2017. “The Mosaic Package: Helping Students to ‘Think with Data’ Using r.” The R Journal 9 (1): 77–102. https://journal.r-project.org/archive/2017/RJ-2017-024/index.html.
Wickham, Hadley, Mara Averick, Jennifer Bryan, Winston Chang, Lucy D’Agostino McGowan, Romain François, Garrett Grolemund, et al. 2019. “Welcome to the tidyverse.” Journal of Open Source Software 4 (43): 1686. https://doi.org/10.21105/joss.01686.