The Great British Visualizations

Proposal-SKAZ

Dataset

The dataset comes from TidyTuesday’s October 25th, 2022 challenge and can be found under the bakeoff package developed by Alison Hill, Chester Ismay, and Richard Iannone. The dataset includes 4 dataframes: bakers, ratings, challenges, and episodes. bakers has 120 rows, each representing an individual baker, and 24 columns. ratings has 94 rows, each representing an episode, and 11 variables, including ratings and the original airdates in the US and the UK. challenges has details about the three types of challenges (signature, technical, and showstopper) as well as the performance of each baker (star baker, eliminated). It has 1136 rows, each representing a baker per episode, and 7 variables. As for episodes, it has 94 rows and 10 variables. Each row represents an episode and the variables describe different attributes of the episode like the number of bakers that appeared in it and the name of eliminated bakers. The variables we plan on using are: percent_episodes_appeared, technical_bottom, technical_top3, series, age, viewers_7day, viewers_28day, uk_airdate, and bakers_out. Respectively, these variables represent the number of episodes in a given series/season in which a given baker appeared out of all episodes aired in that series/season, the number of times a given baker was in the bottom 3 on the technical challenge, the season number (1-10), the age in years in the first episode appeared, the number of viewers in millions within a 7-day window from airdate, the number of viewers in millions within a 28-day window from airdate, the original airdate of an episode in the UK, and the number of bakers were either eliminated, left at will, or left due to illness in that episode.

More information on the dataset can be found under the official reference manual.

We were initially interested in the data set because so many of us enjoyed watching the British and French versions of the show, but our main deciding factor was the fact that the data set has such a vast array of both categorical and numerical variables.

Questions

Question 1

How does a baker’s age (at the time of the competition) and their performance in the technical challenges correlate to the total number of episodes they appear in (i.e. how many episodes they last for)?

Question 2

How do the different characteristics of each season and the progression of episodes within each season influence the show’s viewership?

Analysis plan

Question 1

We will create two scatter plot graphs that have the percentage of episodes the baker appeared in (percent_episodes_appeared) on the y-axis. We’re using the percentage of episodes appeared in rather than the total number of episodes appeared in because although most seasons have 10 episodes, - according to Wikipedia - season 1 has 6 episodes and season 2 has 8 episodes. Because the number of episodes per season isn’t standardized, we can’t equally compare the total number of episodes appeared in across the seasons.

On the x-axis of the first graph we’ll have how many times a baker scored in the bottom three on the technical challenge and how many times a baker scored the absolute lowest on a technical challenge in order to investigate how poor performance in technical challenges correlates to the total number of episodes appeared in.

On the x-axis of the second graph, we’ll have how many times a baker scored in the top three on the technical challenge and how many times a baker scored the absolute highest on a technical challenge in order to investigate how strong performance in technical challenges correlates to the total number of episodes appeared in. Both graphs will have a trend line for each y variable, created using geom_smooth() to better understand the relationship between performance and percentage of episodes appeared in.

Finally, we plan on creating a new variable called age_generation by subtracting the age of the contestants at the time of competition from the air date of the season in order to find the birth year of each contestant. We will then use the birth year to group each contestant by their generation and plot their ages using a discrete color scale. We will be finding the generational divides using the years given by weforum.org. The points will be colored and shaped by age so that we can see if the data of different age groups tends to clump around certain success levels in technical challenges or longevities in the show, which would reflect the priorities of different age groups in the show (do they prioritize technical skills or creativity?) and which age groups tend to last the longest in the show.

Question 2

In order to compare the viewership distribution based on the presenters and network the show was on (which vary by season), we will plot vertical box plots of the 7-day viewership. The 7-day viewership gives an understanding of how many fans of the show are watching it regularly, and while the 28-day viewership would give us a different lens on viewership, the data for 28-day viewership is remarkably inconsistent across seasons (we think there may have been some errors in data collection). The first graph will also be faceted by season to investigate the correlation to viewership of changes in presenters & judges (which changed after season 7) and of changes in the broadcasting network (between seasons the show went through three different broadcasters: BBC Two, BBC One, then Channel 4). We will change the background color of each season’s graph to indicate who the presenters & judges were for the season, and we’ll change the border color of the graph to indicate which network it was on.

For the second plot, we will use a ridge plot in order to investigate how different numbered episodes (for example, episode one of each season vs. episode 6 of each season) correlates to the viewership. Because the show’s relationships (and thus, the show’s drama/intensity) develop as each season progresses, we expect to see the viewership also increase with each episode.