ST558-project-1
COVID-19 API Vignette
C. Donahue June 26, 2022
This document is a vignette to describe how to read and summarize data from an API. I will demonstrate using a COVID-19 API, because my attempts to work with the Pokemon API were unsuccessful. I will build a couple of functions to interact and explore the data we can retrieve from the API.
To use a function that returns data at the country level, the user may either enter the country’s name or its two-letter country code in the ISO 3166 format.
Requirements
I used the following packages in the creation of the vignette, and the user will need them to run the functions and interact with the COVID-19 API:
httr: I used this to access the APIjsonlite: I used this to convert the Pokemon data to a dataframetidyverse: I used this set of packages for things like piping and plotting data visualizationssjmisc: I used this to group a numeric variable into categories to make a histogramcountrycode: I used this to categorize the COVID-19 observations by continent
To get started, install and load the packages listed above:
library(httr)
library(jsonlite)
library(tidyverse)
library(sjmisc)
library(countrycode)
Functions to Interact With the API
The COVID-19 API is simple because it does not require users to get a key and authenticate. This makes accessing it less complex than some other APIs. If you want, you can purchase different monthly subscription levels for more than just the basic authorization to the API. The premium subscriptions include more interesting data (population ages, GDP, diabetes rates, handwashing facilities, etc.) and no rate limit.
countryCases
I wrote a function countryCases for a user to interact with the
summary endpoint of the Coronavirus COVID19 API. It returns a data
frame with key metrics (cases, deaths, recoveries) for every country. It
accepts one argument, country, and the default value is “all”. The
user may enter a country’s name or its two character country code to get
only data for a specific country.
countryCases <- function(country="all"){
###
# This function returns a data frame with data on Covid cases. It can also
# return those columns for a single country if a country's name or abbreviation is passed.
###
# Get the country data from the summaryRoute endpoint.
outputAPI <- fromJSON(
"https://api.covid19api.com/summary"
)
# Select only the Countries data from the JSON output.
output <- outputAPI$Countries
# If country does not equal "all", check if it is an abbrev or country name.
if (country != "all"){
# If country is in the CountryCode column, subset output for just that row.
if (country %in% output$CountryCode){
output <- output %>%
filter(CountryCode == country)
}
# If country is in the Country column, subset output for just that row.
else if (country %in% output$Country){
output <- output %>%
filter(Country == country)
}
# Otherwise, throw an informative error.
else {
message <- paste("ERROR: Argument for country was not found in either",
"the Country or CountryCode columns. Try ('all') to",
"find the country you're looking for.")
stop(message)
}
}
# Do nothing if the country value equals "all".
else {
}
# convert the Date to a Date class
output$Date <- as.Date(output$Date)
# Return the output data frame.
return(output)
}
getSlug
I also wrote a helper function, getSlug, which finds the “slug”, or
the version of a country’s name formatted for use within a URL. It uses
the Countries endpoint. The user inputs a country’s name or two-letter
abbreviation, and getSlug returns the slug. This will help in other
functions that interact with the COVID-19 API.
# Function to get the slug for a given country name or abbreviation
getSlug <- function(country){
# Get the country list from the Countries endpoint.
output <- fromJSON("https://api.covid19api.com/countries")
# If country is in the ISO2 column, subset output for just that row.
if (country %in% output$ISO2){
output <- output$Slug[output$ISO2 == country]
}
# If country is in the Country column, subset output for just that row.
else if (country %in% output$Country){
output <- output$Slug[output$Country == country]
}
# Otherwise, throw an informative error.
else {
message <- paste("ERROR: Argument for country was not found in either",
"the Country or CountryCode columns. Try ('all') to",
"find the country you're looking for.")
stop(message)
}
# Return the slug.
return(output)
}
dailyCases
I wrote a function to allow the user to select a country of interest and
receive the daily confirmed case count for that country, as well as the
country’s location (latitude/longitude). The dailyCases function uses
the helper function getSlug from above to retrieve the information via
the Day One Live endpoint. It returns cumulative confirmed cases
starting with the first day a case was confirmed.
dailyCases <- function(country){
# find the "slug" based on the country the user selected
slug <- getSlug(country)
# Get the country data from the Day One Live endpoint.
output <- fromJSON(paste(
"https://api.covid19api.com/dayone/country/", slug, "/status/confirmed/live", sep = "")
)
# Return the output data frame.
return(output)
}
plotCases
Next, I wanted to create a function to plot the cases by date, so the
user could get a visualization of the type of data they retrieve using
my dailyCases function. The user inputs a country, and receives a
labeled plot of cumulative confirmed cases in that country.
plotCases <- function(country){
data <- dailyCases(country)
# Convert date column from character into date class
data$Date <- as.Date(data$Date)
# Plot the data
ggplot(data, aes(x = Date, y = Cases)) +
geom_col() +
ggtitle(paste("Cumulative Cases by Date in ", country, sep = "")) +
xlab("Date") + ylab("Cumulative Cases")
}
casesByDate
The casesByDate function interacts with the By Country Live
endpoint, using the user’s choice of country, start date, and end date.
It returns a data frame with daily cumulative confirmed case counts, as
well as the country’s name, abbreviation, and location.
# The user inputs a country or 2-letter abbreviation,
# and start/end dates in the format YYYY-MM-DD
casesByDate <- function(country, startDate, endDate){
###
# This function returns a data frame with data on number of cumulative confirmed Covid cases between selected dates in selected country.
###
# find the "slug" based on the country the user selected
slug <- getSlug(country)
# Get the country data from the Day One Live endpoint.
output <- fromJSON(paste(
"https://api.covid19api.com/dayone/country/", slug, "/status/confirmed/live?from=", startDate, "T00:00:00Z&to=", endDate, "T00:00:00Z", sep = "")
)
# Return the output data frame.
return(output)
}
typesByCountry
The typesByCountry function interacts with the
Live By Country and Status endpoint. This function returns a data
frame with data on number of cumulative confirmed Covid cases, deaths,
recoveries, and active cases in the selected country over time.
typesByCountry <- function(country){
# find the "slug" based on the country the user selected
slug <- getSlug(country)
# Get the country data from the Day One Live endpoint.
output <- fromJSON(paste(
"https://api.covid19api.com/dayone/country/", slug, "/status/confirmed", sep = "")
)
# Return the output data frame.
return(output)
}
newCount
After looking through the data while making the functions above, I
wanted to calculate the daily number of new cases, instead of just
looking at cumulative case counts for a country. To get new cases for
today, for example one just needs to subtract yesterday’s cumulative
case count from today’s cumulative case count.
For the newCount function, we interact with the
Live By Country and Status After Date endpoint. (Some other endpoints
do have a count of new cases.) The user parameters include a country of
interest and a start date. The function will output a data frame for
that country with the number of new cases of each type (Confirmed,
Deaths, Recovered, Active), each day.
# The user inputs a country's name or 2-letter abbreviation,
# and start dates in the format "YYYY-MM-DD"
newCount <- function(country, startDate){
# find the "slug" based on the country the user selected
slug <- getSlug(country)
# save the day before so we can do our subtraction to get New cases
# using the cumulative numbers
# Had to actually subtract 2 to get data from the day before
dayBefore <- as.Date(startDate) -2
# Get the country data from the Live By Country And Status After Date endpoint
output <- fromJSON(paste(
"https://api.covid19api.com/live/country/", slug, "/status/confirmed/date/", dayBefore, "T13:13:30Z", sep = "")
)
# Convert the date column to date class
output$Date <- as.Date(output$Date)
# Need to sort by province and then date
# but only if the selected country lists provinces (or states)
if (!all(is.na(output$Province) | output$Province == "")) {
output <- output[order(output$Province),]
}
# needs some work still in order to properly handle countries w/provinces
# Add columns
# for new confirmed
output$newConfirmed[1] <- 0
output$newConfirmed[2:length(output$Confirmed)] <- diff(output$Confirmed[1:length(output$Confirmed)], lag = 1)
# new deaths column
output$newDeaths[1] <- 0
output$newDeaths[2:length(output$Deaths)] <- diff(output$Deaths[1:length(output$Deaths)], lag = 1)
# new recovered
output$newRecovered[1] <- 0
output$newRecovered[2:length(output$Recovered)] <- diff(output$Recovered[1:length(output$Recovered)], lag = 1)
# new active cases
output$newActive[1] <- 0
output$newActive[2:length(output$Active)] <- diff(output$Active[1:length(output$Active)], lag = 1)
# Delete the first observation
# It was only returned initially so we could subtract it to get the new numbers that day
output <- output[-1]
# Return the output data frame.
return(output)
}
getAllCovidData
Finally, I created a function to get all the COVID-19 data from the API
and return a dataframe to the user. This function interacts through the
All Data endpoint, returning 10MB of data, so should be used
infrequently.
getAllCovidData <- function(){
# Get the data from the All Data endpoint.
output <- fromJSON("https://api.covid19api.com/all")
# Return the output data frame.
return(output)
}
Data Exploration
First let’s pull a summary of the COVID-19 data for all countries by
calling getAllCovidData().
covid <- countryCases()
One of the most concerning aspects of COVID-19 is how deadly it is, so I wanted to look at the number of deaths as a percentage of confirmed cases. This variable does not exist, so I need to calculate it. I divided the total number of deaths for each country by the total number of confirmed cases.
# Add a column for the fatality rate.
covid <- covid %>%
mutate(FatalityRate = TotalDeaths / TotalConfirmed)
I would hypothesize that wealthier countries have a lower fatality rate because of access to better medical care and safer living conditions in general. Since this dataset does not give me any measure of a country’s wealth (without a paid subscription), I will just look at a ranking of the best and worst fatality rates to see if I notice anything.
head(covid[order(covid$FatalityRate),], 25)
## ID Country
## 6 c763f496-ce3f-4b06-9e73-889d3d2b8be2 Antarctica
## 74 34bc7c8c-8af2-4f42-9b8e-b90e8cda697e Holy See (Vatican City State)
## 111 00c42d4a-2cf8-4e8e-b515-bd799b660d39 Marshall Islands
## 115 8a19f9de-c22d-47f5-9fa0-8d7685bca242 Micronesia, Federated States of
## 21 f8c5c8b6-a55e-450b-bba5-02bfbfafe4f5 Bhutan
## 77 40ae7a47-795f-4e1c-a6c2-e8caa09f7064 Iceland
## 29 f91ad9cb-a685-4739-b4b9-f726a784fe53 Burundi
## 179 8fc28add-cc67-4de1-81d6-56da73a56ff7 Tonga
## 158 6047d168-e7d9-4575-bce9-9d30b15f80c4 Singapore
## 126 6f13999a-bf93-49e0-9656-451e70bfa139 New Zealand
## 133 7a3e72a8-a47f-4dba-9d29-85809c803d29 Palau
## 10 57cf0389-4ac8-4434-9dd3-67bfe150bf8e Australia
## 190 c9502c22-622a-4356-9269-52563de4899b Vanuatu
## 92 027eef4a-d1d0-4959-b2e7-ca9e8615b7b5 Korea (South)
## 26 6331d499-8b48-485e-8071-49ca4c79b27f Brunei Darussalam
## 108 901a22df-b087-44d2-89b3-5880cce7e418 Maldives
## 173 1a951620-ff5d-41a0-a289-280bf95a0dad Taiwan, Republic of China
## 142 c19f9ac2-c7f6-4b65-9bb3-2aaeaf375f44 Qatar
## 150 c6708fa1-6ab4-4ad4-b911-e5f2f279ab8c Samoa
## 48 ca32a680-fbc2-4e41-82e4-786a83c90779 Denmark
## 45 97364b72-615f-48b8-b3b4-6e1a092cf87e Cyprus
## 130 58163f5a-54d8-48e2-af40-f33795fd1ffb Norway
## 118 cd58b73d-614b-481b-9e87-3e76785ca55f Mongolia
## 14 0d40af08-1009-4a94-ac88-736c2ac202c9 Bahrain
## 185 c985a787-f46c-40a7-b5f7-d5811a8a64eb United Arab Emirates
## CountryCode Slug NewConfirmed TotalConfirmed
## 6 AQ antarctica 0 11
## 74 VA holy-see-vatican-city-state 0 29
## 111 MH marshall-islands 0 18
## 115 FM micronesia 0 38
## 21 BT bhutan 0 59674
## 77 IS iceland 0 192991
## 29 BI burundi 0 42542
## 179 TO tonga 0 12079
## 158 SG singapore 0 1403242
## 126 NZ new-zealand 0 1314155
## 133 PW palau 0 5201
## 10 AU australia 21437 8000312
## 190 VU vanuatu 0 11044
## 92 KR korea-south 0 18326019
## 26 BN brunei 0 159591
## 108 MV maldives 0 180384
## 173 TW taiwan 0 3573703
## 142 QA qatar 0 379277
## 150 WS samoa 0 14812
## 48 DK denmark 0 3207777
## 45 CY cyprus 0 504717
## 130 NO norway 0 1444043
## 118 MN mongolia 0 926282
## 14 BH bahrain 0 616588
## 185 AE united-arab-emirates 0 937037
## NewDeaths TotalDeaths NewRecovered TotalRecovered Date FatalityRate
## 6 0 0 0 0 2022-06-26 0.0000000000
## 74 0 0 0 0 2022-06-26 0.0000000000
## 111 0 0 0 0 2022-06-26 0.0000000000
## 115 0 0 0 0 2022-06-26 0.0000000000
## 21 0 21 0 0 2022-06-26 0.0003519121
## 77 0 153 0 0 2022-06-26 0.0007927831
## 29 0 38 0 0 2022-06-26 0.0008932349
## 179 0 12 0 0 2022-06-26 0.0009934597
## 158 0 1408 0 0 2022-06-26 0.0010033907
## 126 0 1411 0 0 2022-06-26 0.0010736937
## 133 0 6 0 0 2022-06-26 0.0011536243
## 10 27 9682 0 0 2022-06-26 0.0012102028
## 190 0 14 0 0 2022-06-26 0.0012676566
## 92 0 24522 0 0 2022-06-26 0.0013380975
## 26 0 225 0 0 2022-06-26 0.0014098539
## 108 0 300 0 0 2022-06-26 0.0016631187
## 173 0 6120 0 0 2022-06-26 0.0017125094
## 142 0 678 0 0 2022-06-26 0.0017876117
## 150 0 29 0 0 2022-06-26 0.0019578720
## 48 0 6487 0 0 2022-06-26 0.0020222727
## 45 0 1072 0 0 2022-06-26 0.0021239625
## 130 0 3280 0 0 2022-06-26 0.0022714005
## 118 0 2179 0 0 2022-06-26 0.0023524154
## 14 0 1492 0 0 2022-06-26 0.0024197681
## 185 0 2311 0 0 2022-06-26 0.0024662847
What stands out, interestingly, is that a lot of these countries in the top 25 (lowest) fatality rate list are island nations.
tail(covid[order(covid$FatalityRate),], 25)
## ID Country
## 75 301ac9c6-7df9-4c30-ad11-30e79ca5de8c Honduras
## 79 47efc9a4-7a50-424e-b7bc-b6f5b832dd21 Indonesia
## 3 7644a23f-d1cf-405c-982e-74a8502df1cc Algeria
## 35 66a99c76-59ff-4eb5-b99b-c03610ee404b Chad
## 73 e83f57ed-bbf0-4eac-918d-fca2961b3f6f Haiti
## 181 dab3be2f-1b31-4c48-8022-e1b0e26879a9 Tunisia
## 137 cb0a87d1-ea3e-4044-bcd9-42ab16c423b6 Paraguay
## 104 390623c8-6797-4c65-9b30-008a10f17962 Macedonia, Republic of
## 63 74e3c8ec-11dd-42cc-a52c-219710474782 Gambia
## 106 460369d9-819d-4dfe-93da-0a1de477e762 Malawi
## 122 883453fe-15a3-48a2-882c-6f98c4aca58b Myanmar
## 27 49eccb75-f878-4f0c-b9fa-d7924a3aa930 Bulgaria
## 128 97a433c1-75c3-4c88-8db8-33e52ebadd1d Niger
## 99 0f417cf0-d876-4e78-9815-f5bb32a1e300 Liberia
## 52 f6948630-365c-405c-948c-34be9e0ca114 Ecuador
## 23 cba07019-d5c7-4327-9a8b-2e1d3dfc6de6 Bosnia and Herzegovina
## 1 c99c6261-e21b-4ac2-9d21-e14e1e491e17 Afghanistan
## 53 f194ed08-0c9b-4b52-9bb3-3b80e9deb9e5 Egypt
## 162 67bf7221-d9ec-409b-82f2-0c5f27e40fa6 Somalia
## 114 53c22273-c1b7-43d6-b00d-d796ac4dfbce Mexico
## 172 cefdab47-13e3-49b0-b178-4ff5670d62cc Syrian Arab Republic (Syria)
## 138 20f64033-8da9-49bd-a074-fb2566c478ed Peru
## 167 1b48d8b2-1e35-4945-8840-d124dc16e580 Sudan
## 193 e13e8a0a-507e-456f-aa30-dc6cb866aabd Yemen
## 91 9db4fcff-fbc3-45d7-8374-6192e7feb966 Korea (North)
## CountryCode Slug NewConfirmed TotalConfirmed NewDeaths
## 75 HN honduras 0 426490 0
## 79 ID indonesia 0 6078725 0
## 3 DZ algeria 0 266030 0
## 35 TD chad 0 7424 0
## 73 HT haiti 0 31301 0
## 181 TN tunisia 0 1046703 0
## 137 PY paraguay 0 655532 0
## 104 MK macedonia 0 313360 0
## 63 GM gambia 0 12002 0
## 106 MW malawi 0 86348 0
## 122 MM myanmar 0 613553 0
## 27 BG bulgaria 0 1169968 0
## 128 NE niger 0 9031 0
## 99 LR liberia 0 7493 0
## 52 EC ecuador 0 901739 0
## 23 BA bosnia-and-herzegovina 0 378413 0
## 1 AF afghanistan 0 182072 0
## 53 EG egypt 0 515645 0
## 162 SO somalia 0 26748 0
## 114 MX mexico 33646 5956732 65
## 172 SY syria 0 55920 0
## 138 PE peru 2341 3613464 4
## 167 SD sudan 0 62551 0
## 193 YE yemen 0 11824 0
## 91 KP korea-north 0 1 0
## TotalDeaths NewRecovered TotalRecovered Date FatalityRate
## 75 10904 0 0 2022-06-26 0.02556684
## 79 156714 0 0 2022-06-26 0.02578074
## 3 6875 0 0 2022-06-26 0.02584295
## 35 193 0 0 2022-06-26 0.02599677
## 73 837 0 0 2022-06-26 0.02674036
## 181 28670 0 0 2022-06-26 0.02739077
## 137 18963 0 0 2022-06-26 0.02892765
## 104 9322 0 0 2022-06-26 0.02974853
## 63 365 0 0 2022-06-26 0.03041160
## 106 2645 0 0 2022-06-26 0.03063186
## 122 19434 0 0 2022-06-26 0.03167453
## 27 37246 0 0 2022-06-26 0.03183506
## 128 310 0 0 2022-06-26 0.03432621
## 99 294 0 0 2022-06-26 0.03923662
## 52 35705 0 0 2022-06-26 0.03959571
## 23 15799 0 0 2022-06-26 0.04175068
## 1 7717 0 0 2022-06-26 0.04238433
## 53 24722 0 0 2022-06-26 0.04794384
## 162 1361 0 0 2022-06-26 0.05088231
## 114 325576 0 0 2022-06-26 0.05465682
## 172 3150 0 0 2022-06-26 0.05633047
## 138 213447 0 0 2022-06-26 0.05906991
## 167 4951 0 0 2022-06-26 0.07915141
## 193 2149 0 0 2022-06-26 0.18174899
## 91 6 0 0 2022-06-26 6.00000000
The worst fatality rates, meanwhile, seem to include not just generally poorer countries, but many that are also war-torn.
Next, I want to create some charts. I will start with the simple
plotCases function to see what looks interesting. I’m going to choose
Australia, as it’s a developed country but also an island, and then
Canada.
plotCases("Australia")
plotCases("Canada")
Looks like Australia had a lot less gradual increase in case counts, at least compared to Canada. Although both have a large climb right around the beginning of 2022. It is important to note that the scale of the y-axis is not consistent here, so we are really just comparing shape and not scale.
Below I plotted the fatality rate against the total cumulative count of confirmed COVID-19 cases. I added a regression line as well. What really stands out, however, is that there is an outlier (North Korea, in fact) with an incredibly high fatality rate (and another with a really high case count, but not on the same scale).
plot1 <- ggplot(covid, aes(TotalConfirmed,
FatalityRate,
color=FatalityRate)) +
# Add a scatter plot layer and adjust the size and opaqueness of points.
geom_point(size=4, alpha=0.75) +
# Add a color gradient for FatalityRate
scale_color_gradient(low="blue", high="red") +
# Remove the legend because it takes up space.
theme(legend.position="none") +
# Add a black regression line.
geom_smooth(method=lm, formula=y~x, color="black") +
# Add labels to the axes.
scale_x_continuous("Cumulative Confirmed Cases") +
scale_y_continuous("Fatality Rate") +
# Add a title.
ggtitle("Fatality Rate vs. Cumulative Confirmed Cases of COVID-19")
plot1
It is tough to see much here,
since the outlier in red is skewing the y-axis. I wanted to see which
country that was, and also the one skewing the x-axis (to a lesser
degree).
head(covid[order(covid$TotalConfirmed),], 5)
## ID Country
## 91 9db4fcff-fbc3-45d7-8374-6192e7feb966 Korea (North)
## 6 c763f496-ce3f-4b06-9e73-889d3d2b8be2 Antarctica
## 111 00c42d4a-2cf8-4e8e-b515-bd799b660d39 Marshall Islands
## 74 34bc7c8c-8af2-4f42-9b8e-b90e8cda697e Holy See (Vatican City State)
## 115 8a19f9de-c22d-47f5-9fa0-8d7685bca242 Micronesia, Federated States of
## CountryCode Slug NewConfirmed TotalConfirmed
## 91 KP korea-north 0 1
## 6 AQ antarctica 0 11
## 111 MH marshall-islands 0 18
## 74 VA holy-see-vatican-city-state 0 29
## 115 FM micronesia 0 38
## NewDeaths TotalDeaths NewRecovered TotalRecovered Date FatalityRate
## 91 0 6 0 0 2022-06-26 6
## 6 0 0 0 0 2022-06-26 0
## 111 0 0 0 0 2022-06-26 0
## 74 0 0 0 0 2022-06-26 0
## 115 0 0 0 0 2022-06-26 0
tail(covid[order(covid$TotalConfirmed),], 5)
## ID Country CountryCode
## 65 462f89f0-fe8e-4842-96e5-f06813474070 Germany DE
## 61 3fbb471d-ffd2-4f98-98d2-8c0862480743 France FR
## 25 a729488e-840f-49b5-9609-665194b11ddd Brazil BR
## 78 eeb86ea2-4662-463a-a2b9-4cc285c319f5 India IN
## 187 6077ed62-422a-49ae-806b-59273f3e66f3 United States of America US
## Slug NewConfirmed TotalConfirmed NewDeaths TotalDeaths NewRecovered
## 65 germany 1 27771112 0 140734 0
## 61 france 0 30714200 0 150356 0
## 25 brazil 0 32023166 0 670229 0
## 78 india 11739 43389973 25 524999 0
## 187 united-states 39372 86949088 144 1015933 0
## TotalRecovered Date FatalityRate
## 65 0 2022-06-26 0.005067640
## 61 0 2022-06-26 0.004895325
## 25 0 2022-06-26 0.020929505
## 78 0 2022-06-26 0.012099547
## 187 0 2022-06-26 0.011684228
If we rank the data by number of confirmed cases (cumulative), we can see that the outlier in red on the scatter plot is there because North Korea only had one confirmed case, but somehow had six deaths. And the outlier with the incredibly high case count? That’s the United States, with almost 87 million as of late June 2022. I plotted the same thing once more, but left out North Korea.
ggplot(covid[covid$Country != "Korea (North)",], aes(TotalConfirmed,
FatalityRate,
color=FatalityRate)) +
# Add a scatter plot layer and adjust the size and opaqueness of points.
geom_point(size=4, alpha=0.75) +
# Add a color gradient for FatalityRate
scale_color_gradient(low="blue", high="red") +
# Remove the legend because it takes up space.
theme(legend.position="none") +
# Add a black regression line.
geom_smooth(method=lm, formula=y~x, color="black") +
# Add labels to the axes.
scale_x_continuous("Cumulative Confirmed Cases") +
scale_y_continuous("Fatality Rate") +
# Add a title.
ggtitle("Fatality Rate vs. Cumulative Confirmed Cases of COVID-19")
Now the scatter plot looks more reasonable, and it’s easy to see the legitimate outliers (Yemen for Fatality Rate and USA for Cases). And more generally, that countries with higher case counts tend to have not-so-high fatality rates.
I wanted to see how the fatality rate data was shaped or distributed, and whether it was gather around a certain rate, or close to zero. I plotted FatalityRate in a histogram to get an idea. For this, I dropped the North Korea observation, which did not seem to be documented the same way other countries were and was an extreme outlier, causing problems in my visualization.
ggplot(covid[covid$Country != "Korea (North)",], aes(x=FatalityRate)) +
geom_histogram(color = "orange", fill = "pink") +
ggtitle("Histogram of Countries by Fatality Rate") +
ylab("Count of Countries") +
xlab("Deaths per Confirmed Case")
## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
Then I decided to bring in the package countrycode so that I could
categorize the observations by continent.
library(countrycode)
covid$Continent <- countrycode::countrycode(sourcevar = covid[,"Country"],
origin = "country.name",
destination = "continent")
## Warning in countrycode_convert(sourcevar = sourcevar, origin = origin, destination = dest, : Some values were not matched unambiguously: Antarctica, Republic of Kosovo
covid$Continent[covid$Country == "Antarctica"] <- "Antarctica"
covid$Continent[covid$Country == "Republic of Kosovo"] <- "Europe"
Antarctica and Kosovo were the only two observations I had to manually set a continent for. I looked at a contingency table of the new column using `table(covid$Continent) to see how many observations each continent had.
table(covid$Continent)
##
## Africa Americas Antarctica Asia Europe Oceania
## 54 35 1 48 45 12
This took a lot of thought, as the API did not have a lot of data I
could use to make a contingency table, unless I wanted to pay for a
subscription.
Once I had the continents set, I looked at how COVID-19 fatality rates
compared across continents. I looked at means and medians for the total
number of deaths, cases (cumulative), and new cases grouped by
continent.
covid %>%
group_by(Continent) %>%
summarise(n = n(),
meanDeaths = mean(TotalDeaths),
medianDeaths = median(TotalDeaths),
meanCases = mean(TotalConfirmed),
medianCases = median(TotalConfirmed),
meanNewCases = mean(NewConfirmed),
medianNewCases = median(NewConfirmed))
## # A tibble: 6 × 8
## Continent n meanDeaths medianDeaths meanCases medianCases meanNewCases
## <chr> <int> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 Africa 54 4720. 992 222659. 60990. 0
## 2 Americas 35 78758. 5728 4602217. 525716 2540.
## 3 Antarctica 1 0 0 11 11 0
## 4 Asia 48 30087. 8912. 3251606. 859711 676.
## 5 Europe 45 41219 13822 4517730. 1444043 1340.
## 6 Oceania 12 1071. 21.5 791050. 13446. 1786.
## # … with 1 more variable: medianNewCases <dbl>
Excluding Antarctica and Oceania due to the small numbers of countries in each continent, these summaries show that countries in Europe generally have more (reported) COVID-19 deaths than countries in Africa, the Americas, and Asia. I thought that was a little surprising.
Then I looked at the Fatality Rate compared across continents, using a box and whiskers plot. I again dropped the North Korea observation for this plot.
ggplot(covid[covid$Country != "Korea (North)",], aes(Continent, FatalityRate)) +
geom_boxplot(color = "blue", outlier.color = "red") +
ggtitle("Boxplot of COVID-19 Fatality Rate by Continent") +
ylab("Deaths per Confirmed Case")
The boxplot was interesting, and could probably be moreso if I had split the “Americas” continent category up by North and South America. You can see that the fatality rates in Oceania are significantly lower than those in Africa and the Americas, because the boxes to not overlap vertically. But I also noticed that the fatality rate for Europe was lower than Africa and the Americas. So the higher death toll in Europe noted in the summary above could be due to the higher case count there, not because the disease was more deadly.
A lot of American tourists are starting to travel overseas this summer
after taking a couple years off. Let’s look at data for some common
destinations. I will use newCount() to get data sets for Austria,
Switzerland, and Greece from the beginning of May 2022.
austria <- newCount("Austria", "2022-05-01")
switzerland <- newCount("Switzerland", "2022-05-01")
greece <- newCount("Greece", "2022-05-01")
Let’s plot the daily count of new cases for these countries. First, I am going to join the data sets by the date column, though.
vacation <- rbind(austria, switzerland, greece)
ggplot(vacation, aes(fill=Country, y= newConfirmed, x=Date)) +
geom_bar(position = "dodge", stat = "identity") +
ggtitle("Summer 2022 New COVID-19 Cases by Country") +
ylab("New Cases")
This plot is not as compelling as I had hoped because it appears that Switzerland is only reporting new cases every week or so, and Austria is not doing so on the weekends. These varying reporting timelines mean visualizing the daily data does not necessarily give the whole picture or make for great comparison. Still, from this chart, it does seem that from the beginning of May until the end of June, new case counts are generally increasing in these three summer travel destinations.
Finally, desperate to create another numerical summary, I thought I would see how COVID looked across months of the year. I chose another fun vacation destination, Portugal, and looked at new cases since the summer of 2021 when the COVID vaccines began taking effect in a lot of developed countries.
portugal <- newCount("Portugal", "2021-06-01")
monthlyNewPortugal <- portugal %>%
group_by(month = lubridate::floor_date(Date, "month")) %>%
summarise(monthlyNewCases = sum(newConfirmed))
monthlyNewPortugal
## # A tibble: 13 × 2
## month monthlyNewCases
## <date> <dbl>
## 1 2021-06-01 7316
## 2 2021-07-01 88846
## 3 2021-08-01 69978
## 4 2021-09-01 32511
## 5 2021-10-01 21358
## 6 2021-11-01 54454
## 7 2021-12-01 214475
## 8 2022-01-01 1253069
## 9 2022-02-01 646523
## 10 2022-03-01 275594
## 11 2022-04-01 307218
## 12 2022-05-01 790879
## 13 2022-06-01 488870
ggplot(monthlyNewPortugal, aes(x= month, y= monthlyNewCases, fill="pink" )) +
geom_bar(stat = "identity") +
ggtitle("New COVID-19 Cases by Month in Portugal") +
xlab("Month") +
ylab("New Cases") +
theme(legend.position = "none")
It looks like, despite vaccines, Portugal was hit pretty hard with new COVID cases around the time the Omicron variant became widespread last winter.
Wrap-Up
To summarize everything I did in this vignette, I built functions to interact with some of the COVID-19 API’s endpoints, retrieved some of the data, and explored it using tables, numerical summaries, and data visualization. I found some interesting things, like that some island nations had more success keeping COVID cases at bay for longer and keeping the COVID mortality rate low, while people living in poorer or war-torn countries had a harder time with surviving the pandemic.
I hope this vignette helps my readers more successfully interact with APIs. It has been good practice for me.