When Not in Rome…

…still do as the Romans do. The Roman Empire built many amphitheaters outside of its capital. This post explores 268 of these historic sites and includes a dashboard for interactive exploration.

R
EDA
tables
maps
interactive
Author

Christian Gebhard

Published

July 8, 2022

Photo by Tomas Robertson on Unsplash

Introduction

Roman amphitheaters are monumental historic buildings, dating back to the antique times of the Roman Empire. They were mainly used for entertainment, hosting gladiator combats or venationes (animal hunts).

On Amphitheaters

One of the best known amphitheatres is the Colosseum in Rome, also known as the “Flavian Amphitheater”. But over several centuries, the Romans built many more across their Empire. The name describes the architecture: the spectator seats (théatron) are arranged around or on both sides (amphi) of the arena in a circular or oval manner.

Data Source

The dataset comprises historic and geospacial data on 268 theaters1.

Acknowledgements

The data was composed and published by Sebastian Heath from the INSTITUTE FOR THE STUDY OF THE ANCIENT WORLD at NYU. Thanks and credits go to Sebastian Heath, as he published the data under the “Unlicense”, which allowed me to explore and analyse the set for this post.

I stumbled upon this set in the great Data is Plural Newsletter by Jeremy Singer-Vine.

Further Sources

For this post I read articles in several online resources, including

Packages

Code
library(tidyverse, quietly = TRUE)
library(crosstalk)
library(leaflet)
library(reactable)
library(reactablefmtr)
library(ggdist)
library(ggiraph)
library(ggtext)
library(MetBrewer)
library(geomtextpath)

library(showtext)
font_add_google("Open Sans")
font_add_google("Bitter")
showtext_auto()
#> ─ Session info ───────────────────────────────────────────────────────────────
#>  setting  value
#>  version  R version 4.2.1 (2022-06-23)
#>  os       Ubuntu 20.04.5 LTS
#>  system   x86_64, linux-gnu
#>  ui       X11
#>  language (EN)
#>  collate  de_DE.UTF-8
#>  ctype    de_DE.UTF-8
#>  tz       Europe/Berlin
#>  date     2022-09-18
#>  pandoc   2.14.2 @ /usr/bin/ (via rmarkdown)
#>  quarto   1.1.251 @ /opt/quarto/bin/quarto
#> 
#> ─ Packages ───────────────────────────────────────────────────────────────────
#>  package       * version    date (UTC) lib source
#>  crosstalk     * 1.2.0      2021-11-04 [1] CRAN (R 4.2.0)
#>  dplyr         * 1.0.9      2022-04-28 [1] CRAN (R 4.2.0)
#>  forcats       * 0.5.2      2022-08-19 [1] CRAN (R 4.2.1)
#>  geomtextpath  * 0.1.0.9000 2022-07-07 [1] Github (AllanCameron/geomtextpath@f11e256)
#>  ggdist        * 3.2.0      2022-07-19 [1] CRAN (R 4.2.1)
#>  ggiraph       * 0.8.3      2022-08-19 [1] CRAN (R 4.2.1)
#>  ggplot2       * 3.3.6      2022-05-03 [1] CRAN (R 4.2.0)
#>  ggtext        * 0.1.1      2020-12-17 [1] CRAN (R 4.2.1)
#>  leaflet       * 2.1.1      2022-03-23 [1] CRAN (R 4.2.0)
#>  MetBrewer     * 0.2.0      2022-03-21 [1] CRAN (R 4.2.0)
#>  purrr         * 0.3.4      2020-04-17 [3] RSPM (R 4.2.0)
#>  reactable     * 0.3.0      2022-05-26 [1] CRAN (R 4.2.0)
#>  reactablefmtr * 2.1.0      2022-06-05 [1] Github (kcuilla/reactablefmtr@ca67199)
#>  readr         * 2.1.2      2022-01-30 [1] CRAN (R 4.2.0)
#>  sessioninfo   * 1.2.2      2021-12-06 [1] CRAN (R 4.2.0)
#>  showtext      * 0.9-5      2022-02-09 [1] CRAN (R 4.2.0)
#>  showtextdb    * 3.0        2020-06-04 [1] CRAN (R 4.2.0)
#>  stringr       * 1.4.1      2022-08-20 [1] CRAN (R 4.2.1)
#>  sysfonts      * 0.8.8      2022-03-13 [1] CRAN (R 4.2.0)
#>  tibble        * 3.1.8      2022-07-22 [1] CRAN (R 4.2.1)
#>  tidyr         * 1.2.0      2022-02-01 [1] CRAN (R 4.2.0)
#>  tidyverse     * 1.3.2      2022-07-18 [3] RSPM (R 4.2.0)
#> 
#>  [1] /home/christian/R/x86_64-pc-linux-gnu-library/4.2
#>  [2] /usr/local/lib/R/site-library
#>  [3] /usr/lib/R/site-library
#>  [4] /usr/lib/R/library
#> 
#> ──────────────────────────────────────────────────────────────────────────────

Exploratory Data Analysis

Next, let’s read the actual amphitheater data and have a look at it.

Code
# read data and drop columns that won't be used
amphi <- readr::read_csv("https://raw.githubusercontent.com/roman-amphitheaters/roman-amphitheaters/d1b2cb2b401e583cc13837451ed403b42e8fceae/roman-amphitheaters.csv") |> 
  select(
    title, label, 
    pleiades, buildingtype, 
    chronogroup, capacity, 
    modcountry, 
    arenamajor, arenaminor, 
    extmajor, extminor, 
    longitude, latitude, elevation)

If you want to see more than the summary, check out the code and output below deck. In there I cover extreme values, distribution of variables and check for spurious correlations.

EDA Summary

There are 268 entries in total and I selected 14 columns of interest.

Missing Data

There are no missing values for the name and location data including coordinates and in which modern country the arena is located now. Other interesting measurements do have missing data unfortunately:

  • external theater measurements: 96 missing (35.8%)
  • arena measurements: 116 missing (43.3%)
  • spectator capacity: 139 missing (51.9%)

Extreme Values

The lowest amphitheater is located in today’s Israel at -134m, the highest at 1170m in Algeria. The one furthest north is located in Newstead (UK), the arena furthest south at Eleutheropolis (Israel).

Below Deck

The following steps were performed to check the validity of the dataset. As this stays below deck, I used base R plots and default colors mostly.

Get an idea of the data

Code
dplyr::glimpse(amphi)
#> Rows: 268
#> Columns: 14
#> $ title        <chr> "Amphitheater at Dura Europos", "Amphitheater at Arles", …
#> $ label        <chr> "Dura", "Arles", "Lyon", "Ludus Magnus", "Colosseum", "Am…
#> $ pleiades     <chr> "https://pleiades.stoa.org/places/893989", "https://pleia…
#> $ buildingtype <chr> "amphitheater", "amphitheater", "amphitheater", "practice…
#> $ chronogroup  <chr> "severan", "flavian", "second-century", "imperial", "flav…
#> $ capacity     <dbl> 1000, 23354, 20000, NA, 50000, 7000, 3500, 22000, 15000, …
#> $ modcountry   <chr> "Syria", "France", "France", "Italy", "Italy", "Italy", "…
#> $ arenamajor   <dbl> 31.0, 47.0, 67.6, NA, 83.0, NA, 47.0, 66.0, 64.0, 37.0, 5…
#> $ arenaminor   <dbl> 25.0, 32.0, 42.0, NA, 48.0, NA, 38.0, 35.0, 41.0, 23.0, 4…
#> $ extmajor     <dbl> 50.0, 136.0, 105.0, NA, 189.0, 88.0, 71.0, 135.0, 126.0, …
#> $ extminor     <dbl> 44.0, 107.0, NA, NA, 156.0, 75.8, 56.0, 104.0, 102.0, 60.…
#> $ longitude    <dbl> 40.728926, 4.631111, 4.830556, 12.494913, 12.492269, 12.5…
#> $ latitude     <dbl> 34.74985, 43.67778, 45.77056, 41.88995, 41.89017, 41.8877…
#> $ elevation    <dbl> 223, 21, 206, 22, 22, 48, 253, 21, 231, 83, 100, 19, 41, …
Code
head(amphi)
#> # A tibble: 6 × 14
#>   title    label pleia…¹ build…² chron…³ capac…⁴ modco…⁵ arena…⁶ arena…⁷ extma…⁸
#>   <chr>    <chr> <chr>   <chr>   <chr>     <dbl> <chr>     <dbl>   <dbl>   <dbl>
#> 1 Amphith… Dura  https:… amphit… severan    1000 Syria      31        25      50
#> 2 Amphith… Arles https:… amphit… flavian   23354 France     47        32     136
#> 3 Amphith… Lyon  https:… amphit… second…   20000 France     67.6      42     105
#> 4 Ludus M… Ludu… https:… practi… imperi…      NA Italy      NA        NA      NA
#> 5 Flavian… Colo… https:… amphit… flavian   50000 Italy      83        48     189
#> 6 Amphith… Amph… https:… amphit… severan    7000 Italy      NA        NA      88
#> # … with 4 more variables: extminor <dbl>, longitude <dbl>, latitude <dbl>,
#> #   elevation <dbl>, and abbreviated variable names ¹​pleiades, ²​buildingtype,
#> #   ³​chronogroup, ⁴​capacity, ⁵​modcountry, ⁶​arenamajor, ⁷​arenaminor, ⁸​extmajor
Code
tail(amphi)
#> # A tibble: 6 × 14
#>   title    label pleia…¹ build…² chron…³ capac…⁴ modco…⁵ arena…⁶ arena…⁷ extma…⁸
#>   <chr>    <chr> <chr>   <chr>   <chr>     <dbl> <chr>     <dbl>   <dbl>   <dbl>
#> 1 Amphith… Tren… https:… amphit… second…      NA Italy        NA      NA      NA
#> 2 Amphith… Aven… https:… amphit… second…   13006 Switze…      51      39      99
#> 3 Amphith… Vena… https:… amphit… imperi…   15000 Italy        60      35     110
#> 4 Amphith… Sain… <NA>    amphit… first-…    3000 France       54      30      65
#> 5 Amphith… Tole… https:… amphit… imperi…      NA Spain        NA      NA      NA
#> 6 Amphith… Kais… https:… amphit… fourth…      NA Switze…      NA      NA      50
#> # … with 4 more variables: extminor <dbl>, longitude <dbl>, latitude <dbl>,
#> #   elevation <dbl>, and abbreviated variable names ¹​pleiades, ²​buildingtype,
#> #   ³​chronogroup, ⁴​capacity, ⁵​modcountry, ⁶​arenamajor, ⁷​arenaminor, ⁸​extmajor
Code
summary(amphi)
#>     title              label             pleiades         buildingtype      
#>  Length:268         Length:268         Length:268         Length:268        
#>  Class :character   Class :character   Class :character   Class :character  
#>  Mode  :character   Mode  :character   Mode  :character   Mode  :character  
#>                                                                             
#>                                                                             
#>                                                                             
#>                                                                             
#>  chronogroup           capacity      modcountry          arenamajor    
#>  Length:268         Min.   : 1000   Length:268         Min.   : 25.00  
#>  Class :character   1st Qu.: 5150   Class :character   1st Qu.: 47.50  
#>  Mode  :character   Median :10000   Mode  :character   Median : 58.00  
#>                     Mean   :12100                      Mean   : 57.18  
#>                     3rd Qu.:15550                      3rd Qu.: 67.00  
#>                     Max.   :50000                      Max.   :101.00  
#>                     NA's   :139                        NA's   :115     
#>    arenaminor       extmajor         extminor        longitude     
#>  Min.   :19.00   Min.   : 39.60   Min.   : 34.00   Min.   :-8.493  
#>  1st Qu.:32.92   1st Qu.: 75.50   1st Qu.: 58.95   1st Qu.: 5.326  
#>  Median :38.75   Median : 95.00   Median : 75.00   Median :10.890  
#>  Mean   :38.03   Mean   : 97.15   Mean   : 76.92   Mean   :10.567  
#>  3rd Qu.:43.00   3rd Qu.:115.75   3rd Qu.: 94.00   3rd Qu.:14.184  
#>  Max.   :62.00   Max.   :189.00   Max.   :156.00   Max.   :40.729  
#>  NA's   :116     NA's   :81       NA's   :96                       
#>     latitude       elevation      
#>  Min.   :31.61   Min.   :-121.00  
#>  1st Qu.:38.48   1st Qu.:  34.75  
#>  Median :42.09   Median : 121.00  
#>  Mean   :42.25   Mean   : 196.79  
#>  3rd Qu.:45.60   3rd Qu.: 286.25  
#>  Max.   :55.60   Max.   :1170.00  
#>
Code
dplyr::count(amphi, buildingtype, sort = TRUE)
#> # A tibble: 6 × 2
#>   buildingtype                 n
#>   <chr>                    <int>
#> 1 amphitheater               255
#> 2 gallo-roman-amphitheater     6
#> 3 practice-arena               3
#> 4 oval-structure               2
#> 5 arena-in-hippodrome          1
#> 6 arena-in-stadium             1
Code
dplyr::count(amphi, chronogroup, sort = TRUE)
#> # A tibble: 18 × 2
#>    chronogroup                         n
#>    <chr>                           <int>
#>  1 imperial                          103
#>  2 second-century                     54
#>  3 flavian                            24
#>  4 first-century                      18
#>  5 republican                         17
#>  6 julio-claudian                     15
#>  7 hadrianic                           7
#>  8 severan                             6
#>  9 augustan                            4
#> 10 caesarean                           4
#> 11 late-second-century                 3
#> 12 third-century                       3
#> 13 fourth-century                      2
#> 14 late-first-century                  2
#> 15 late-first-early-second-century     2
#> 16 post-severan                        2
#> 17 neronian                            1
#> 18 trajanic                            1
Code
dplyr::count(amphi, modcountry, sort = TRUE)
#> # A tibble: 25 × 2
#>    modcountry         n
#>    <chr>          <int>
#>  1 Italy            105
#>  2 France            36
#>  3 Tunisia           29
#>  4 Spain             15
#>  5 United Kingdom    15
#>  6 Algeria            8
#>  7 Switzerland        7
#>  8 Turkey             7
#>  9 Austria            6
#> 10 Germany            5
#> # … with 15 more rows

Distribution of numeric variables

Code
hist(amphi$capacity)

Code
hist(amphi$arenamajor)

Code
hist(amphi$arenaminor)

Code
hist(amphi$extmajor)

Code
hist(amphi$extminor)

Code
hist(amphi$elevation)

Extreme Values

One value caught my eye: the lowest elevation is more than 100m below sea level, which seems odd on first thought. A quick lookup in pleiades and wikipedia however confirms, that the Roman theater of Scythopolis in today’s ‘Beit She’an’ lies below sea level within the Jordan Rift Valley.

The highest located amphitheater is located in today’s Algeria, called ‘Amphitheater at Lambaesis’.

Correlation patterns

Most of the following variable correlations do not make sense in the real world, but this is intended to check for spurious correlations. The strong correlations of external measurements, arena measurements and capacity seem quite plausible.

Code
# select numeric columns
amphi.num <- dplyr::select_if(amphi, is.numeric)

# calculate correlation matrix
amphi.corr <- cor(
  amphi.num,
  use = "pairwise.complete.obs"
)

# plot correlation matrix
corrplot::corrplot(amphi.corr, "circle")

There is a slight negative correlation between the elevation and the theater measurements, which I cannot explain at this time. To check for visually apparent patterns, we’ll add a scatterplot matrix including the columns that have a Spearman’s \(\rho > 0.1\).

Code
amphi.num |> 
  select(-c(longitude, latitude, arenaminor)) |> 
  plot()

External and Internal Measures of the Amphitheaters

Next up is an analysis of the size of the theaters. Available in the dataset are outer measures and arena size. The amphitheaters usually were of oval shape, so there is a longest possible and a shortest possible axis. Another measure is the capacity of spectators, which will be looked at later.

The buildings and arenas were not always circles. For the calculation of the area we’ll assume, that the shapes are perfect ellipses2.

As preliminary step I derived several variables from the existing columns, such as area and measurements relative to the Colosseum in Rome. The values were stored in amphi.measures. Check out the code below deck, if you like.

Summary

The amphitheater with the largest arena area is located at Utica in Tunisia (the area is given in \(m^2\)). The Colosseum, officially called the “Flavian Amphitheater at Rome”, ranks on place 6 in this category:

Code
amphi.measures |> 
  arrange(desc(arenaarea)) |> 
  head() |> 
  select(title, arenaarea, modcountry)
#> # A tibble: 6 × 3
#>   title                                              arenaarea modcountry 
#>   <chr>                                                  <dbl> <chr>      
#> 1 Amphitheater at Utica                                  3770. Tunisia    
#> 2 Amphitheater at Altinum                                3644. Italy      
#> 3 Amphitheater at Octodurus/Forum Claudii Vallensium     3603. Switzerland
#> 4 Amphitheater at Caesarea                               3490. Israel     
#> 5 Amphitheater at Lucca                                  3330. Italy      
#> 6 Flavian Amphitheater at Rome                           3129. Italy

On the other hand, the Colosseum could – by far – harbor the largest audience:

Code
amphi.measures |> 
  arrange(desc(capacity)) |> 
  head() |> 
  select(title, capacity, modcountry)
#> # A tibble: 6 × 3
#>   title                            capacity modcountry
#>   <chr>                               <dbl> <chr>     
#> 1 Flavian Amphitheater at Rome        50000 Italy     
#> 2 Imperial Amphitheater at Capua      37000 Italy     
#> 3 Flavian Amphitheater at Pozzuoli    35700 Italy     
#> 4 Amphitheater at Thysdrus            35000 Tunisia   
#> 5 Amphitheater at Tours               34000 France    
#> 6 Amphitheater at Milan               31649 Italy

To visualize how many people could see an event in the Colosseum, compared to the other venues, we’ll plot the distribution in a raincloud plot. The majority of theaters lie between 5000 to 20000 visitors.

Code
p <- amphi.measures |> 
  mutate(
    is_colosseum = label == "Colosseum",
    psize = ifelse(is_colosseum, 3, 0.5)
  ) |> 
    ggplot() +
  aes(x=1, y = capacity) +
  ggdist::stat_halfeye(
    fill = "#845d29",
    width = .2, 
    .width = 0, 
    justification = -2.5, 
    point_colour = NA,
    alpha = 0.85) + 
  ggdist::stat_pointinterval(
    color = "black",
    position = position_nudge(x = 0.45),
  ) +
  geom_point_interactive(
    aes(tooltip = title, color = is_colosseum, size = psize),
    # size = 2,
    alpha = .4,
    position = position_jitter(
      seed = 753, width = .4
    )
  ) +
  coord_flip() +
  scale_color_met_d("Isfahan1") +
  theme_classic() +
  labs(
    title = "Visitor Capacity of Roman Amphitheaters",
    subtitle = "The <span style='color:#178f92; weight: bold;'>Colosseum in Rome</span> is the largest venue with 50k seats.<br>The majority of theaters could fit between 5k and 20k spectators.",
    y = "Visitor capacity",
    caption = "dataviz by @c_gebhard on jollydata.blog | 2022<br>Data by Sebastian Heath, Institute for the Study of the Ancient World, NYU"
  ) +
  theme(
    axis.line.y = element_blank(),
    axis.title.y = element_blank(),
    axis.text.y = element_blank(),
    axis.ticks = element_blank(),
    panel.grid.major.x = element_line(color = "#DDDDDD"),
    plot.title = element_markdown(family = "Bitter", size = 12, face = "bold"),
    plot.subtitle = element_markdown(size = 10),
    plot.caption = element_markdown(family = "Bitter", size = 8, lineheight = 1.2),
    legend.position = "none"
  )

girafe(
  ggobj = p,
  height_svg = 4
  )