Data Visualization

Kable package

In a previous weeks, we saw R Markdown in action, where multiple things can be created in one location: code, commentary, and output.

In this chapter we will explore package which will facilitate the creation of presentation-worthy tables: “kableExtra”.

Let’s work with the cross-sectional data on the credit history for a sample of applicants for a type of credit card.

data(CreditCard)
cardHead <- head(CreditCard)
cardHead

##   card reports      age income       share expenditure owner selfemp dependents
## 1  yes       0 37.66667 4.5200 0.033269910  124.983300   yes      no          3
## 2  yes       0 33.25000 2.4200 0.005216942    9.854167    no      no          3
## 3  yes       0 33.66667 4.5000 0.004155556   15.000000   yes      no          4
## 4  yes       0 30.50000 2.5400 0.065213780  137.869200    no      no          0
## 5  yes       0 32.16667 9.7867 0.067050590  546.503300   yes      no          2
## 6  yes       0 23.25000 2.5000 0.044438400   91.996670    no      no          0
##   months majorcards active
## 1     54          1     12
## 2     34          1     13
## 3     58          1      5
## 4     25          1      7
## 5     64          1      5
## 6     54          1      1

cardHead %>%
  kbl()

card	reports	age	income	share	expenditure	owner	selfemp	dependents	months	majorcards	active
yes	0	37.66667	4.5200	0.0332699	124.983300	yes	no	3	54	1	12
yes	0	33.25000	2.4200	0.0052169	9.854167	no	no	3	34	1	13
yes	0	33.66667	4.5000	0.0041556	15.000000	yes	no	4	58	1	5
yes	0	30.50000	2.5400	0.0652138	137.869200	no	no	0	25	1	7
yes	0	32.16667	9.7867	0.0670506	546.503300	yes	no	2	64	1	5
yes	0	23.25000	2.5000	0.0444384	91.996670	no	no	0	54	1	1

Let’s tweak the appearance of this with the “align” and the “caption” arguments.

The align argument takes a character vector with letters “l”, “c”, or “r” - specifying where you want the columns to be aligned.

The caption argument gives a caption to the table.

base <- cardHead %>%
  kbl(align = c(rep("c", 7), rep("r", 5)), caption = "kable example with card data")
base

kable example with card data

card	reports	age	income	share	expenditure	owner	selfemp	dependents	months	majorcards	active
yes	0	37.66667	4.5200	0.0332699	124.983300	yes	no	3	54	1	12
yes	0	33.25000	2.4200	0.0052169	9.854167	no	no	3	34	1	13
yes	0	33.66667	4.5000	0.0041556	15.000000	yes	no	4	58	1	5
yes	0	30.50000	2.5400	0.0652138	137.869200	no	no	0	25	1	7
yes	0	32.16667	9.7867	0.0670506	546.503300	yes	no	2	64	1	5
yes	0	23.25000	2.5000	0.0444384	91.996670	no	no	0	54	1	1

A key function, where we can enjoy much of the configuration for the table, is via kable_styling().

We have options “bootstrap_options” or “latex_options”, where the latter requires the use of the package “tinytex” and a local installation of LaTeX.

Possible options for “bootstrap_options” include ‘basic’, ‘striped’, ‘bordered’, ‘hover’, ‘condensed’, ‘responsive’, and none.

Possible for “latex_options” include ‘basic’, ‘striped’, ‘hold_position’, ‘HOLD_position’, ‘scale_down’, and ‘repeat_header’.

base %>%
  kable_styling(bootstrap_options = "striped")

kable example with card data

card	reports	age	income	share	expenditure	owner	selfemp	dependents	months	majorcards	active
yes	0	37.66667	4.5200	0.0332699	124.983300	yes	no	3	54	1	12
yes	0	33.25000	2.4200	0.0052169	9.854167	no	no	3	34	1	13
yes	0	33.66667	4.5000	0.0041556	15.000000	yes	no	4	58	1	5
yes	0	30.50000	2.5400	0.0652138	137.869200	no	no	0	25	1	7
yes	0	32.16667	9.7867	0.0670506	546.503300	yes	no	2	64	1	5
yes	0	23.25000	2.5000	0.0444384	91.996670	no	no	0	54	1	1

Next, we can customize the look and feel of particular rows and columns.

Let’s see an example here, where we make the last three rows blue.

base %>%
  kable_styling(bootstrap_options = "bordered") %>%
  column_spec(8:12, bold = T) %>%
  row_spec(4:6, italic = T, color = "gold", background = "blue")

kable example with card data

card	reports	age	income	share	expenditure	owner	selfemp	dependents	months	majorcards	active
yes	0	37.66667	4.5200	0.0332699	124.983300	yes	no	3	54	1	12
yes	0	33.25000	2.4200	0.0052169	9.854167	no	no	3	34	1	13
yes	0	33.66667	4.5000	0.0041556	15.000000	yes	no	4	58	1	5
yes	0	30.50000	2.5400	0.0652138	137.869200	no	no	0	25	1	7
yes	0	32.16667	9.7867	0.0670506	546.503300	yes	no	2	64	1	5
yes	0	23.25000	2.5000	0.0444384	91.996670	no	no	0	54	1	1

We can also create groups for our columns.

base %>%
  kable_styling(bootstrap_options = "bordered") %>%
  add_header_above(c("Group 1" = 4, "Group 2" = 2, "Group 3" = 6))

kable example with card data

Group 1				Group 2		Group 3
card	reports	age	income	share	expenditure	owner	selfemp	dependents	months	majorcards	active
yes	0	37.66667	4.5200	0.0332699	124.983300	yes	no	3	54	1	12
yes	0	33.25000	2.4200	0.0052169	9.854167	no	no	3	34	1	13
yes	0	33.66667	4.5000	0.0041556	15.000000	yes	no	4	58	1	5
yes	0	30.50000	2.5400	0.0652138	137.869200	no	no	0	25	1	7
yes	0	32.16667	9.7867	0.0670506	546.503300	yes	no	2	64	1	5
yes	0	23.25000	2.5000	0.0444384	91.996670	no	no	0	54	1	1

Data Aggregation

In the first stage of our analysis we are going to group our data in the form of the simple frequency table.

First, let’s take a look at the distribution of income in our sample and verify the tabular accuracy using TAI measure:

options(scipen=999)

limits<- cut(CreditCard$income,seq(0,14,by=2))

Without ‘kable’ styling it’s quite ugly right? ;-)

Tabular accuracy

An index of tabular accuracy TAI, described by Jenks and Casspal in 1971 is to optimize the class distribution used in a cartograms/frequency tables etc.

The TAI indicator takes values in the range (0;1). The numerator of the expression is the sum of the absolute deviations of the values classified into classes, and the denominator is the sum of the absolute deviations of the entire classified set.

The better the class division reflects the nature of the data, the larger the indicator will be. As the number of classes increases, the indicator will take on larger values.

Let’s calculate TAI index to check the properties of the tabulated data:

tabelka2 <- classIntervals(CreditCard$income, n=7, style="fixed", fixedBreaks=seq(0,14,by=2))
jenks.tests(tabelka2)

##        # classes  Goodness of fit Tabular accuracy 
##        7.0000000        0.9085328        0.6568085

As we can see - TAI index…

We can use different recipes… (styles):

tabelka3<-classIntervals(CreditCard$income, n=10, style="sd")
plot(tabelka3,pal=c(1:10))

jenks.tests(tabelka3)

##        # classes  Goodness of fit Tabular accuracy 
##        8.0000000        0.9274792        0.6909392

Still, the TAI indicator is not satisfactory. What should we change in the final frequency table design?

hist(CreditCard$income)

Continuous variables

We can calculate the absolute and relative frequencies of a vector x with the function ‘Freq’ from the DescTools packages. Continuous (numeric) variables will be cut using the same logic as used by the function hist. Categorical variables will be aggregated by table. The result will contain single and cumulative frequencies for both, absolute values and percentages.

tabela4<-Freq(CreditCard$income,breaks=seq(0,14,by=2),useNA="ifany")

tabela4 %>%
  kable(col.names = c("Incomes in kUSD","Frequency","Percentage %","Cumulative frequency","Cumulative percentage %")) %>%
  kable_classic(full_width = F, html_font = "Cambria")

Incomes in kUSD	Frequency	Percentage %	Cumulative frequency	Cumulative percentage %
[0,2]	236	0.1789234	236	0.1789234
(2,4]	783	0.5936315	1019	0.7725550
(4,6]	205	0.1554208	1224	0.9279757
(6,8]	63	0.0477635	1287	0.9757392
(8,10]	23	0.0174375	1310	0.9931766
(10,12]	7	0.0053071	1317	0.9984837
(12,14]	2	0.0015163	1319	1.0000000

BTW: what about TAI of that table?…

Categorical variables

Now, let’s take a look at the categorical data and make some tabulations. The xtabs function works like table except it can produce tables from frequencies using the formula interface.

Let’s say we want to see the table with data on how many card applications was accepted or not:

## card
##   no  yes 
##  296 1023

We may easily produce cross-tabs (status vs. Does the individual own their home?) as well:

crosstab<-xtabs(~ card + owner, data=CreditCard)
crosstab

##      owner
## card   no yes
##   no  206  90
##   yes 532 491

and transform it into pretty html table with the kable function:

crosstab %>% 
  kbl() %>%
  kable_styling(full_width = F) %>%
  column_spec(1, bold = T, border_right = T) %>%
  column_spec(2, background = "yellow")

	no	yes
no	206	90
yes	532	491

Data Visualization

We will explore the “ggplot2” package of the tidyverse for data visualization purposes. The “ggplot2” packages involve the the following three mandatory components:

1. Data
2. An aesthetic mapping
3. Geoms (aka objects)

The following components can also optionally be added:

4. Stats (aka transformations)
5. Scales
6. Facets
7. Coordinate systems
8. Position adjustments
9. Themes

Please note that code in this tutorial was adapted from Chapters 3 of the book “R for Data Science” by Hadley Wickham and Garrett Grolemund.

The full book can be found at: https://r4ds.had.co.nz/

A good cheat sheet for ggplot2 functions can be found at: https://rstudio.com/wp-content/uploads/2015/03/ggplot2-cheatsheet.pdf

Scatterplots

Let’s create an extremely simple scatterplot.

We will use the function ggplot() to do this.

The format of any ggplot graph is this function, followed by another function to add objects.

The objects on a graph in the case of a scatterplot are points. The function we add to it is geom_point.

These functions rely on a function on the inside called aes().

The data and aesthetic mapping components can be added to either the ggplot() or geom functions.

ggplot(data = mpg) +
  geom_point(aes(x = displ, y = hwy))

This is one of the most basic graphs that one can make using the ggplot2 framework.

Next, let’s add color.

geom_point() understands the following aesthetics: x, y, alpha, color, fill, group, shape, size, and stroke (see help documentation).

Let’s map the color argument to the variable “class” from mpg.

ggplot(mpg) +
  geom_point(aes(x = displ, y = hwy, color = class))

This is not the only way to color objects.

Including the color argument inside of the aes() function can map colors to a choice of variable.

However, we can specify colors manually, by specifying color outside of the aes() function. We will also illustrate the “size” argument.

ggplot(mpg) +
  geom_point(aes(x = displ, y = hwy, size = class), color = "blue")

Barplots

Lastly, let’s examine other objects that we can plot using ggplot(). We will create a bar chart using the function geom_bar().

ggplot(mpg) +
  geom_bar(aes(x = class))

With the geom_bar() function, we have a great use-case for a stat transformation.

The following code can be used to convert these counts to proportions:

ggplot(mpg) +
  geom_bar(aes(x = class, y = stat(prop), group = 1))

Histograms

Next, let’s create a histogram with the geom_histogram() function.

ggplot(mpg) +
  geom_histogram(aes(x = hwy))

The geom_histogram() function accepts the argument “binwidth”, and has two key arguments for color: fill (this controls the overall color), and color (this controls the border).

Let’s fill all these in:

ggplot(mpg) +
  geom_histogram(aes(x = hwy), binwidth = 5, fill = "navy", color = "gold")

geom_histogram() provides a great example to modify the scale.

Notice in this example that the axis is automatically broken up by units of 10, and does not begin at 0.

We can modify this with the function scale_x_continuous(), as well as the y-axis with the function scale_y_continuous().

There are three key arguments we will feed this function: “breaks”, “limits”, and “expand”.

“breaks” will define the breaks on the axis.

“limits” will define the beginning and end of the axis, and the “expand” argument can be used to start the axes at 0 by using “expand = c(0,0)”.

ggplot(mpg) +
  geom_histogram(aes(x = hwy), binwidth = 5, fill = "navy", color = "gold") +
  scale_x_continuous(breaks = seq(0, 45, 5), limits = c(0, 50), expand = c(0,0)) +
  scale_y_continuous(breaks = seq(0, 90, 10), limits = c(0, 90), expand = c(0,0))

Boxplots

Next, we will create boxplots.

p <- ggplot(mpg) +
  geom_boxplot(aes(x = class, y = cty, fill = class))
p

Facets

Faceting generates small multiples each showing a different subset of the data. Small multiples are a powerful tool for exploratory data analysis: you can rapidly compare patterns in different parts of the data and see whether they are the same or different.

Read more about facets here.

Notice in this document the use of the fig.height and fig.width options.

Key arguments to facet_wrap() are “facets”, “nrow”, and “ncol”.

ggplot(mpg) +
  geom_boxplot(aes(x = class, y = cty, fill = class)) +
  facet_wrap(facets = ~cyl, nrow = 2, ncol = 2)

Coordinates

Other coordinate systems can be applied to graphs created from ggplot2.

One example is coord_polar(), which uses polar coordinates. Most of these are quite rare. Probably the most common one is coord_flip(), which will flip the X and Y axes. Let’s also illustrate the labs() function, which can be used to change labels.

ggplot(mpg) +
  geom_bar(aes(x = class, fill = factor(cyl))) +
  labs(title = "Cylinders by Class", fill = "cylinders") +
  coord_flip()

These bars are stacked on top of each of other, due to the “cyl” variable being mapped to the “fill” argument. There are various position adjustments that can be used. Again, most of these are not very common, but a common one is the argument “position = ‘dodge’”, which will put items side-by-side.

See this example:

ggplot(mpg) +
  geom_bar(aes(x = class, fill = factor(cyl)), position = "dodge") +
  labs(title = "Cylinders by Class", fill = "cylinders") + 
  coord_flip()

Themes

Lastly, we can alter the “theme”, or the overall appearance of our plot.

I recommend using the ggThemeAssist package, because this will make this incredibly easy, with an interface that will automatically generate reproducible code.

This can be used by highlighting a ggplot2 object, and navigating to Addins > ggplot Theme Assistant.

We’ll make the following changes: eliminating the panel grid lines, eliminating axis ticks, adding a title called “Boxplot Example”, making it bigger and putting it in bold, and adjusting it to the center.

# p
p + theme(axis.ticks = element_line(linetype = "blank"),
    panel.grid.major = element_line(linetype = "blank"),
    panel.grid.minor = element_line(linetype = "blank"),
    plot.title = element_text(size = 14,
        face = "bold", hjust = 0.5)) +labs(title = "Boxplot Example")

There are many more examples of things that can be done with ggplot2.

It is an amazingly powerful and flexible package, and it is worth getting acquainted with the cheat sheet.

Exercise 1.

Using data on credit card applications’ status please present the frequency table with the nice, kable format for average monthly credit card expenditures of applicants.

# In order to create monthly average we just divide the expenditure column by the months column. We also don't count the applicants which value in 'months' column equals to '0' in order to avoid dividing by '0'. We also create a frequency table showing the frequency of applicants in each interval of the Monthly Credit Card Expenditure.  


frequency_table <- CreditCard %>%
   filter(months != 0) %>% 
   mutate(AverageMonthlyExpenditure = expenditure / months) %>%
   arrange(desc(AverageMonthlyExpenditure))

top_ten_rows <- head(frequency_table, 10)

kable_table <- top_ten_rows %>%
  kable("html", caption = "Top 10 Average Monthly Credit Card Expenditures") %>%
  kable_styling(bootstrap_options = c("striped", "hover"))

kable_table

Top 10 Average Monthly Credit Card Expenditures

	card	reports	age	income	share	expenditure	owner	selfemp	dependents	months	majorcards	active	AverageMonthlyExpenditure
472	yes	1	43.33333	4.80	0.3706127	1482.3680	yes	no	0	3	1	4	494.1227
765	yes	0	44.50000	9.00	0.2599927	1949.8620	no	no	2	4	1	7	487.4655
624	yes	0	27.16667	2.60	0.8778461	1902.0000	no	no	0	4	1	7	475.5000
1292	yes	0	22.00000	1.75	0.6214691	906.3092	no	no	0	2	1	0	453.1546
1181	yes	0	23.16667	2.60	0.8245508	1786.2770	no	no	0	4	1	0	446.5693
874	yes	0	25.58333	2.80	0.3252597	758.9391	no	no	0	2	1	7	379.4696
434	yes	0	28.91667	2.75	0.3822785	875.8884	yes	no	1	3	1	3	291.9628
487	yes	0	22.91667	2.50	0.2395248	498.9267	no	no	0	2	1	0	249.4633
763	yes	0	35.25000	6.00	0.4582348	2291.1740	no	no	2	10	1	0	229.1174
113	yes	0	28.16667	2.40	0.1137675	227.3683	no	no	0	1	1	29	227.3683

interval_data <- CreditCard %>%
   filter(months != 0) %>% 
   mutate(AverageMonthlyExpenditure = expenditure / months) %>%
   mutate(ExpendituresPerMonthIntervals = cut(AverageMonthlyExpenditure, breaks = seq(0, max(AverageMonthlyExpenditure,      na.rm = TRUE) + 100, by = 100), right = FALSE)) %>%
   group_by(ExpendituresPerMonthIntervals) %>%
   summarise(Frequency = n(), .groups = 'drop') %>%
   arrange(desc(Frequency))

kable_table_intervals <- interval_data %>%
  kable("html", caption = "Frequency of Applicants by Monthly Credit Card Expenditure") %>%
  kable_styling(bootstrap_options = c("striped", "hover"))

kable_table_intervals

Frequency of Applicants by Monthly Credit Card Expenditure

ExpendituresPerMonthIntervals	Frequency
[0,100)	1275
[100,200)	26
[200,300)	9
[400,500)	5
[300,400)	1

Exercise 2.

The data comes from https://flixgem.com/ (dataset version as of March 12, 2021). The data contains information on 9425 movies and series available on Netlix.

Answer with the most appropriate data visualization for the following questions:

1. What is the distribution of Imdb scores for Polish movies and movie-series?

# To select only Polish Productions we grep the only one which have a "Polish" tag in their Tags column

polish_films <- mydata[grep("Polish", mydata$Tags), ]

ggplot(polish_films, aes(x = IMDb.Score)) +
  geom_histogram(binwidth = 0.05, fill = "orange", color = "black") +
  labs(title = "Distribution of IMDb Scores for Polish Productions",
       x = "IMDb Score",
       y = "Frequency")

2. What is the density function of Imdb scores for Polish movies and movie-series?

# The structure is really similar we just exchange the histogram to density chart
polish_films <- mydata[grep("Polish", mydata$Tags), ]

ggplot(polish_films, aes(x = IMDb.Score)) +
  geom_density(fill = "steelblue", alpha = 0.9) +
  labs(title = "Density of IMDb Scores for Polish Productions",
       x = "IMDb Score",
       y = "Density") +
  theme_minimal()

3. What are the most popular languages available on Netflix?

# Using separate_rows method we split the rows of films with more than language in Languages column and then we sum the occurrences. We only display 10 the most popular ones in order for the plot to look nice and clear 

library(tidyr)

expanded_data <- mydata %>%
  separate_rows(Languages, sep = ",\\s*")

language_counts <- expanded_data %>%
  group_by(Languages) %>%
  summarise(count = n()) %>%
  arrange(desc(count)) %>%
  top_n(10, count)

ggplot(language_counts, aes(x = reorder(Languages, -count), y = count, fill = Languages)) +
  geom_bar(stat = "identity") +
  labs(title = "Most Popular Languages",
       x = "Languages",
       y = "Frequency")

For extra credits:

Extra challenge 1.: Create a chart showing actors starring in the most popular productions.

# In order to choose the actors who played in the most popular productions we used the Awards.Nominated.For column. 
# Assuming that the production with huge number of nominations must be popular, we sum up the nominations for a film for each actor that played in it.

mydata_separated_actors <- mydata %>%
  separate_rows(Actors, sep = ",\\s*")

actor_awards_counted <- mydata_separated_actors %>%
  group_by(Actors) %>%
  summarise(TotalNominations = sum(Awards.Nominated.For, na.rm = TRUE)) %>%
  arrange(desc(TotalNominations))

top_actors <- head(actor_awards_counted, 10)

ggplot(top_actors, aes(x = reorder(Actors, -TotalNominations), y = TotalNominations, fill = Actors)) +
  geom_bar(stat = "identity") +
  labs(title = "Top 10 Actors by Award Nominations",
       x = "Actors",
       y = "Total Award Nominations")  +
  theme_minimal() +
  theme(axis.text.x = element_text(angle = 45, hjust = 1))

Extra challenge 2.: For movies and series, create rating charts from the various portals (Hidden Gem, IMDb, Rotten Tomatoes, Metacritic). Hint: it’s a good idea to reshape the data to long format.

# In order to get more clarity we created two charts, one with "Hidden.Gem.Score" and "IMDb.Score" and then second one with "Rotten.Tomatoes.Score", "Metacritic.Score". That's due to max score being 10 in the HiddenGem abd IMDb and 100 in RottenTomatoes and Metacritic. 

movies_long <- mydata %>%
  pivot_longer(
    cols = c(Hidden.Gem.Score, IMDb.Score, Rotten.Tomatoes.Score, Metacritic.Score), 
    names_to = "RatingSource", 
    values_to = "Score"
  ) %>%
  na.omit()

ratings_0_10 <- movies_long %>%
  filter(RatingSource %in% c("Hidden.Gem.Score", "IMDb.Score"))

ratings_0_100 <- movies_long %>%
  filter(RatingSource %in% c("Rotten.Tomatoes.Score", "Metacritic.Score"))

#First plot
ggplot(ratings_0_10, aes(x = RatingSource, y = Score, fill = RatingSource)) +
  geom_boxplot() +
  labs(title = "Ratings from 0 to 10 (Hidden Gem and IMDb)",
       x = "Rating Source",
       y = "Rating Score") +
  theme_minimal() +
  theme(legend.position = "none")  

#Second plot
ggplot(ratings_0_100, aes(x = RatingSource, y = Score, fill = RatingSource)) +
  geom_boxplot() +
  labs(title = "Ratings from 0 to 100 (Rotten Tomatoes and Metacritic)",
       x = "Rating Source",
       y = "Rating Score") +
  theme_minimal() +
  theme(legend.position = "none")

Extra challenge 3.: Which film studios produce the most and how has this changed over the years?

# Firstly we omit the productions that are not produced by any film studio. Then we group the result by release year and production house. Again for more clarity of the chart, we choose only 5 the most popular studios. Then we plot the result using geom_line.
# Also to show how much overall every Production House has created films and TV series we just group by the Production.House counting the appearances, of course beforehandly separating the Production.House column

movies <- mydata %>%
  filter(!is.na(Production.House) & Production.House != "")  %>%  
  mutate(Release.Date = mdy(Release.Date),  
         Release.Year = year(Release.Date)) %>%
  separate_rows(Production.House, sep = ",\\s*")  

production_counts <- movies %>%
  group_by(Production.House, Release.Year) %>%
  summarise(Films_Produced = n(), .groups = 'drop')  

top_production_houses <- production_counts %>%
  group_by(Production.House) %>%
  summarise(Total_Films = sum(Films_Produced)) %>%
  top_n(5, Total_Films) %>%
  pull(Production.House)

filtered_production_counts <- production_counts %>%
  filter(Production.House %in% top_production_houses)

ggplot(filtered_production_counts, aes(x = Release.Year, y = Films_Produced, color = Production.House)) +
  geom_line() +
  scale_x_continuous(limits = c(1960, 2024)) +
  labs(title = "Film Production Over the Years by Top Studios",
       x = "Year",
       y = "Number of Films Produced") +
  theme_minimal()

production_totals <- mydata %>%
  filter(!is.na(Production.House) & Production.House != "") %>%
  separate_rows(Production.House, sep = ",\\s*") %>%
  group_by(Production.House) %>%
  summarise(Total_Productions = n(), .groups = 'drop') %>%
  arrange(desc(Total_Productions))

top_production_houses <- production_totals %>%
  slice_max(order_by = Total_Productions, n = 10)

ggplot(top_production_houses, aes(x = reorder(Production.House, Total_Productions), y = Total_Productions, fill = Production.House)) +
  geom_bar(stat = "identity") +
  coord_flip() +  # Flipping coordinates for better readability of production house names
  labs(title = "Total Productions by Top 10 Production Houses",
       x = "Production House",
       y = "Total Productions") +
  theme(legend.position = "none")