#==============================================================================# # Scooby-Doo # #==============================================================================# # The data used this week is collected on ScoobyPedia, an encyclopedia on the # hit cartoon series Scooby-Doo. This is a quote from their website: # # The show follows the iconic mystery solving detectives, know as Mystery Inc., # as they set out to solve crime and unmask criminals, bent on revenge or # committing criminal acts for their own personal gain. # # Titular character, Scooby, is followed by his best pal Shaggy as both vie for # Scooby Snacks on their adventures! Velma brings her extra intellect and # initiative to them, setting out plans to catch criminals. Fred is the team's # leader while Daphne is bold and full of personality. # The data we will be using has been coded for research purposes and # demonstrated by the Tidy Tuesday Initiative. The data is described here: # https://github.com/rfordatascience/tidytuesday/blob/master/data/2021/2021-07-13/readme.md # Package loading -------------------------------------------------------------- library(ggplot2) # For plotting library(readr) # For loading data library(dplyr) # For tidy wrangling library(stringr) # For string operations # Data loading ----------------------------------------------------------------- # We can load the data directly from Github. scooby_doo <- read_csv( "https://raw.githubusercontent.com/rfordatascience/tidytuesday/master/data/2021/2021-07-13/scoobydoo.csv") # Glimpse the data structure: scooby_doo %>% glimpse() # Q1: What can we learn about the data from glimpsing it? What is the unit of # observation? Are all variables coded properly? How can we fix it? # Practice tidy data wrangling: filtering and selecting ------------------------ # Say we only want to use data from the original series "Scooby Doo, Where Are # You!". We can subset the data by using filter(). # Exercise A: create a new dataset called scooby_doo_orig by filtering the original dataset. # Q2: How many episode did the original series have? # For a later exercise, we wish to subset the dataset of the original series to # include each episode's index, title, and monster data. # Exercise B: Create this dataset and call it scooby_doo_orig_monster. Check # that you have 9 columns in the new dataset. # Average IMDB rating per season, per series ----------------------------------- # For this, we will use the full dataset again. # Exercise C: inspect the number of episodes in each season per series (hint: try # grouping the appropriate variables and counting the resulting number of cases) # Inspecting the results a little tells us that there are some special and # crossover episodes from series that are not Scooby Doo-centered. There are # also a few movies. Say we are only interested in regular Scooby Doo episodes. # Exercise D: Create a new dataset called scooby_doo_ep that only include Scooby # Doo TV Series and segmented TV series. # Exercise E: Summarize the average IMDB rating per season of each # series. # Q3: What percentage of the real culprits were unsuspected? ----------------------- # We can find this out by summing up the number of "TRUE" in the column # non_suspect. We can then compare this number to the number of culprits who # were suspected (marked by "FALSE"). # Exercise F: Find out the percentage of real culprits that were unsuspected. # Q4: Who caught the most culprits? ------------------------------------------------ # To answer this question, we need to count the number of "TRUE" in each column # containing the term "caught". # Step 1: Select columns containing the term "caught": caught <- scooby_doo_ep %>% select(starts_with("caught")) glimpse(caught) # We have a dataset containing one column per character, with a logical value # for each episode (caught/not). By summing up each column, we can find out the # number of times "TRUE" is coded for each character. # Step 2: Sum up each column to find the number of "TRUE" incidences caught_total <- caught %>% colSums(na.rm = TRUE) %>% # Results in a named vector enframe() # Nicely transforms the named vector to a dataframe. caught_total # Step 3: Sort caught_total in a descending order to easily find the best monster # catcher. caught_total %>% arrange(value %>% desc()) # It looks like Scooby is our champion! # Q5: Who unmasked the most monsters? ---------------------------------------------- # Exercise G: Repeat the steps above with the appropriate variables to find out. # Q6: Who was captured the most times? --------------------------------------------- # Exercise H: Find out! # Q7: Can we break a myth? ----------------------------------------------------- # Scooby-Doo is known for unmasking many monsters to be fake. Of the mythical # monsters, how many were real? # Let's first inspect the different categories under the variable monster_type: scooby_doo_ep %>% group_by(monster_type) %>% count() # We quickly notice that some cells are coded with multiple monster_type values. # That is because the dataset is structured in a wide format, with one episode # per row. If we wish to analyze the different mosters, we need to pivot the # dataset to a longer format, analyzing one monster per row. # Note: For simplicity, we will only focus on monster_type, although more # variables are structured the same way. scooby_doo_ep_longtype <- scooby_doo_ep %>% separate_longer_delim(cols = monster_type, delim = ",") scooby_doo_ep_longtype %>% View() # Exercise I: filter the long dataset to select "real"monsters only and count # the number of incidences for each monster type left. # Data visualization ----------------------------------------------------------- # For GGPlot, we always have to specify an aesthetic argument, where we specify # what should be treated as the x, y, or group variable. After we specify that, # we can add new arguments by using the "+" sign and adding more and more layers # to the plot. # Plot 1: IMDB Rating per date, for different formats ========================= # For example, if we want to produce a point-plot, we can do so with: scooby_doo %>% ggplot(aes(x = date_aired, y = imdb, color = format) ) + geom_point() # Plot 2: Suspected vs unsuspected culprits in different formats =============== # For a bar chart, we only need to specify an x or a y variable: scooby_doo %>% ggplot(aes(x = non_suspect)) + geom_bar() # Plot 3: Who caught the most culprits? ======================================== # We can plot a summary table like the one we created for "caught_total" with a # geom_col() command. # Let's remember first how the summary table looked like head(caught_total) # And plot the summary table caught_total %>% ggplot(aes(x = name, y = value)) + geom_col() # Plot 4: Per monster type, how many real vs unreal monsters were uncovered? ======================================= # Earlier, we saw how we can create a bar chart with ggplot. But what if we want # to add another layer to the data, for example different categories? We can do # so by adding an aesthetic "fill". # Note: Fill refers to the color of the chart, while "color" refers to its # border. Therefore, for some chart types (e.g. point, line), we should specify # a "color" aesthetic instead of "fill" for illustrating different categories. # Try both on the code below! scooby_doo_ep_longtype %>% ggplot(aes(y = monster_type, fill = monster_real) ) + geom_bar(position = "dodge") # Try removing the position argument