Hello! I’ve been learning to code in R Studio and I’ve done this project to show you a bit of what I’ve been learning regarding data visualization.
This project is about the evolution of Covid 19 in Peru from March 6, 2020 to March 28, 2021. We will make 4 graphs. The first 3 graphs correspond to the number of confirmed, recovered and deaths daily cases that have occurred in Peru. These 3 graphs will be made using the dygraph library.
On the other hand, the 4th graph will be made with the plotly library and we will use the accumulated values of confirmed cases, deaths and recovered in the same period of time as the previous plots.
Finally, we will merge these 4 graphics into a single figure using combineWidgets from the manipulateWidget library.
Let’s start by calling our libraries
Let’s call the datasets from the CSSE-John Hopkins University with the command read.csv
confirmed <-read.csv("/Users/marcoarellano/Desktop/DATA SCIENCE/COVID 19/03.28.2021/DATA/time_series_covid19_confirmed_global.csv")
deaths<-read.csv("/Users/marcoarellano/Desktop/DATA SCIENCE/COVID 19/03.28.2021/DATA/time_series_covid19_deaths_global.csv")
recovered <-read.csv("/Users/marcoarellano/Desktop/DATA SCIENCE/COVID 19/03.28.2021/DATA/time_series_covid19_recovered_global.csv")Datasets come in wide format and, by default, have an X in front of each column name. To fix this, we use the substring command. This command allows us to delete characters from the names of the columns that we want to eliminate, in this case the X.
names(deaths)[5:396] <- substring(names(deaths)[5:396],2)
names(recovered)[5:396] <- substring(names(recovered)[5:396],2)
names(confirmed)[5:396] <- substring(names(confirmed)[5:396],2)Now we convert our 3 datasets from wide format to long format using pivot_longer.
Additionally for future analysis and projects we add a new variable, continents, which corresponds to the list of continents to which each country in our datasets belongs.
Finally, the variable dates which corresponds to the new variable that we created by joining all the columns of the wide format. We convert it to a date format using lubridate::mdy.
continents <- read_excel("~/Desktop/DATA SCIENCE/COVID 19/03.28.2021/DATA/continents_Corrected.xlsx")
confirmed_long <- confirmed %>%
inner_join(continents, by = "Country.Region") %>%
pivot_longer (
cols = !c(Province.State , Country.Region , Lat, Long , continent , region),
names_to = c("dates"),
values_to = "confirmed")%>%
mutate(dates= mdy(dates))%>%
group_by(Country.Region, continent, region, dates)%>%
summarise(confirmed= sum(confirmed))%>%
ungroup()## `summarise()` has grouped output by 'Country.Region', 'continent', 'region'. You can override using the `.groups` argument.deaths_long <- deaths %>%
inner_join(continents, by = "Country.Region") %>%
pivot_longer (
cols = !c(Province.State , Country.Region , Lat, Long , continent , region),
names_to = c("dates"),
values_to = "deaths")%>%
mutate(dates= mdy(dates))%>%
group_by(Country.Region, continent, region, dates)%>%
summarise(deaths= sum(deaths))%>%
ungroup()## `summarise()` has grouped output by 'Country.Region', 'continent', 'region'. You can override using the `.groups` argument.recovered_long <- recovered %>%
inner_join(continents, by = "Country.Region") %>%
pivot_longer (
cols = !c(Province.State , Country.Region , Lat, Long , continent , region),
names_to = c("dates"),
values_to = "recovered")%>%
mutate(dates= mdy(dates))%>%
group_by(Country.Region, continent, region, dates)%>%
summarise(recovered= sum(recovered))%>%
ungroup()## `summarise()` has grouped output by 'Country.Region', 'continent', 'region'. You can override using the `.groups` argument.As we already have our 3 datasets in the desired formats, we create a new column that corresponds to the number of daily cases.
Daily cases are achieved by subtracting the cases from day 1 to day 2, the difference give us the case increase in a single day. To achieve this, we use the lag command that allows us to find the number of cases in the day before.
confirmed_long <- confirmed_long %>%
arrange(dates) %>%
group_by(Country.Region) %>%
mutate(confirmed_dailycases = confirmed - lag(confirmed, default = 0)) %>%
ungroup()
deaths_long <- deaths_long %>%
arrange(dates) %>%
group_by(Country.Region) %>%
mutate(deaths_dailycases = deaths - lag(deaths, default = 0)) %>%
ungroup()
recovered_long <- recovered_long %>%
arrange(dates) %>%
group_by(Country.Region) %>%
mutate(recovered_dailycases = recovered - lag(recovered, default = 0)) %>%
ungroup()Now let’s create a working dataset for the country Peru.
First we select the contry Peru in our 3 datasets and combine them together in a single dataset using full_join.
Additionally, we create two new variables, deaths_rate and deaths_100k, for future use.
Peru_confirmed<- confirmed_long%>%
filter(Country.Region %in% "Peru")%>%
select(Country.Region , dates, confirmed , confirmed_dailycases)
Peru_confirmed <- Peru_confirmed[,-c(1)]
Peru_deaths<- deaths_long%>%
filter(Country.Region %in% "Peru")%>%
select(Country.Region , dates, deaths , deaths_dailycases)
Peru_deaths <- Peru_deaths[,-c(1)]
Peru_recovered<- recovered_long%>%
filter(Country.Region %in% "Peru")%>%
select(Country.Region , dates, recovered, recovered_dailycases)
Peru_recovered <- Peru_recovered[,-c(1)]
Peru_global<- Peru_confirmed %>%
full_join(Peru_deaths , by= "dates")%>%
full_join(Peru_recovered , by= "dates")%>%
mutate(deaths_rate= round ((deaths/confirmed)*100, 1 ),
deaths_100k= ceiling((deaths/32625948)*10^5))Next, we create a 7-day Rolling Average variable for confirmed, deaths and recovered variables. The position of our 7-day Rolling Average corresponds to day 4 of the 7-day interval. The 7-day Rolling Averages smooth the day to day observed variation in 7 consecutive days.
nr<- nrow(Peru_global)
Peru_global$rid<- seq(1, nr ,1)
Peru_global_ts <- as_tsibble(Peru_global,
key= rid,
index = dates)
Peru_global_ts <- Peru_global_ts %>%
filter_index("2020-03-06" ~ .) %>%
mutate(confirmed7_dailycases= slide_index_dbl(.i = dates,
.x = confirmed_dailycases,
.f = mean,
.before = 3,
.after= 3),
deaths7_dailycases = slide_index_dbl(.i = dates,
.x = deaths_dailycases,
.f = mean,
.before = 3,
.after= 3),
recovered7_dailycases = slide_index_dbl(.i = dates,
.x = recovered_dailycases,
.f = mean,
.before = 3,
.after= 3),
confirmed7= slide_index_dbl(.i = dates,
.x = confirmed,
.f = mean,
.before = 3,
.after= 3),
deaths7 = slide_index_dbl(.i = dates,
.x = deaths,
.f = mean,
.before = 3,
.after= 3),
recovered7 = slide_index_dbl(.i = dates,
.x = recovered,
.f = mean,
.before = 3,
.after= 3))Now let’s graph! We use the dygraph library to graph an interactive time series of confirmed, recovered and deaths daily cases. Each plot will have 2 variables: the daily number of cases and 7-day moving average. The interactive graph allows us to zoom in selected time intervals for a more detailed view of the series.
peru_int_confirmeddaily <- cbind(Peru_global_ts[c(1,3,11)])
ts_peru_int_confirmeddaily<- xts(x = peru_int_confirmeddaily,
order.by =peru_int_confirmeddaily$dates)
ts_peru_int_confirmeddaily <- ts_peru_int_confirmeddaily[,-c(1)]
dygraph(ts_peru_int_confirmeddaily,
main = "Confirmed Covid 19 Daily cases") %>%
dySeries("confirmed_dailycases", stepPlot = TRUE,
fillGraph = TRUE, color = "lightblue" , label = "Confirmed Daily Cases" )%>%
dySeries("confirmed7_dailycases", drawPoints = TRUE,
pointShape = "square", color = "darkblue" , label = "Moving Avg 7") %>%
dyRangeSelector(height = 20)%>%
dyLegend(width = 300)peru_int_recovereddaily <-cbind(Peru_global_ts[c(1,7,13)])
ts_peru_int_recovereddaily<- xts(x = peru_int_recovereddaily,
order.by = peru_int_recovereddaily$dates)
ts_peru_int_recovereddaily <- ts_peru_int_recovereddaily[,-c(1)]
dygraph(ts_peru_int_recovereddaily,
main = " Recovered Covid 19 Daily cases") %>%
dySeries("recovered_dailycases", stepPlot = TRUE,
fillGraph = TRUE, color = "turquoise" , label = "Recovered Daily Cases") %>%
dySeries("recovered7_dailycases", drawPoints = TRUE,
pointShape = "circle", color = "green" , label = "Moving Avg 7")%>%
dyRangeSelector(height = 20)%>%
dyLegend(width = 300)peru_int_deathsdaily <-cbind(Peru_global_ts[c(1,5,12)])
ts_peru_int_deathsdaily<- xts(x = peru_int_deathsdaily,
order.by = peru_int_deathsdaily$dates)
ts_peru_int_deathsdaily <- ts_peru_int_deathsdaily[,-c(1)]
dygraph(ts_peru_int_deathsdaily,
main = "Covid 19 Daily Deaths") %>%
dySeries("deaths_dailycases", stepPlot = TRUE, fillGraph = TRUE,
color = "orange" , label = "Deaths Daily Cases") %>%
dySeries("deaths7_dailycases", drawPoints = TRUE, pointShape = "square",
color = "red" , label = "Moving Avg 7")%>%
dyRangeSelector(height = 20)%>%
dyLegend(width = 285)The plotly library is used to create the last graph. For this graph we use the accumulated values and the 7 days moving averages of our 3 variables: confirmed, deaths and recovered.
This chart will have 2 y-axis. The y-axis on the left corresponds to the values of confirmed and recovered cases; on the other hand, the right y-axis corresponds to the values of deaths. We considered to have two y-axis because the confirmed and recovered values have a similar range, in contrast to the death values that had a lower range. For that reason, in order to visualize the trend in a better way, it was decided to add the second y-axis.
knitr::opts_chunk$set(echo = TRUE, fig.align="left")
plot_ly() %>%
add_trace(x = ~Peru_global_ts$dates, y = ~ round(Peru_global_ts$confirmed7,1), name = "Confirmed",
type = 'scatter', mode = 'lines',line = list(color = 'blue', size= 4),
hoverinfo = "text",
text = ~paste("Date: ", Peru_global_ts$dates,
"<br>",
"Confirmed: ", round(Peru_global_ts$confirmed7,1)))%>%
add_trace(x = ~Peru_global_ts$dates, y = ~round(Peru_global_ts$recovered7,1), name = "Recovered",
type = 'scatter', mode = 'lines',line = list(color = 'green', size= 4),
hoverinfo = "text",
text = ~paste("Date: ",Peru_global_ts$dates,
"<br>",
"Recovered: ",round(Peru_global_ts$recovered7,1)))%>%
add_trace(x = ~Peru_global_ts$dates, y = ~round(Peru_global_ts$deaths7,1), name = "Deaths", yaxis = "y2",
type = 'scatter', mode = 'lines',line = list(color = 'red', size= 4),
hoverinfo = "text",
text = ~paste("Date: ",Peru_global_ts$dates,
"<br>",
"Deaths: ",round(Peru_global_ts$deaths7,1))) %>%
layout( title = list(text="Cumulative Covid 19 cases Peru 2020-2021",size= 10),
yaxis2 = list(tickfont =list(color = "red"),
overlaying = "y",
side = "right",
title = "Cumulative Deaths",
showgrid= FALSE),
xaxis = list(title="Dates",
color="black"),
yaxis= list(tickangle=0,
title="Cumulative Confirmed and Recovered <br><br><br>",
standoff= 90,
showgrid= FALSE),
legend = list(orientation = "h",
xanchor = "center",
x = 0.5,
y=-0.1),
autosize = T,
margin = list( l = 100, r = 100, b = 100, t = 100, pad = 20))After we have all our graphics ready, we combine them in a single figure that has two columns; the left column will have a combined cumulative cases series graph, and the right column will have three individual graphs corresponding to daily cases.
We use the command manipulateWidget::combineWidget. This command allows us to join our interactive graphics in a single image in a quick and easy way.
First, we create a function to combine in a single graph the three cumulative series.
cumulates_plotly<- function(id){plot_ly() %>%
add_trace(x = ~Peru_global_ts$dates, y = ~ round(Peru_global_ts$confirmed7,1), name = "Confirmed",
type = 'scatter', mode = 'lines',line = list(color = 'blue', size= 4),
hoverinfo = "text",
text = ~paste("Date: ", Peru_global_ts$dates,
"<br>",
"Confirmed: ", round(Peru_global_ts$confirmed7,1)))%>%
add_trace(x = ~Peru_global_ts$dates, y = ~round(Peru_global_ts$recovered7,1), name = "Recovered",
type = 'scatter', mode = 'lines',line = list(color = 'green', size= 4),
hoverinfo = "text",
text = ~paste("Date: ",Peru_global_ts$dates,
"<br>",
"Recovered: ",round(Peru_global_ts$recovered7,1)))%>%
add_trace(x = ~Peru_global_ts$dates, y = ~round(Peru_global_ts$deaths7,1), name = "Deaths", yaxis = "y2",
type = 'scatter', mode = 'lines',line = list(color = 'red', size= 4),
hoverinfo = "text",
text = ~paste("Date: ",Peru_global_ts$dates,
"<br>",
"Deaths: ",round(Peru_global_ts$deaths7,1))) %>%
layout( title = list(text="Cumulative Covid 19 cases Peru 2020-2021",size= 10),
yaxis2 = list(tickfont =list(color = "red"),
overlaying = "y",
side = "right",
title = "Cumulative Deaths",
showgrid= FALSE,
mirror=FALSE),
xaxis = list(title="Dates",
color="black"),
yaxis= list(tickangle=0,
title="Cumulative Confirmed and Recovered <br><br><br>",
standoff= 90,
showgrid= FALSE,
mirror=FALSE),
legend = list(orientation = "h",
xanchor = "center",
x = 0.5,
y=-0.2),
autosize = T,
margin = list( l = 100, r = 100, b = 100, t = 100, pad = 20))}Second, we create a function to define each component of the right column in the figure.
c1 <-function(id){dygraph(ts_peru_int_confirmeddaily,
main = "Confirmed Covid 19 Daily cases") %>%
dySeries("confirmed_dailycases", stepPlot = TRUE,
fillGraph = TRUE, color = "lightblue" , label = "Confirmed Daily Cases" )%>%
dySeries("confirmed7_dailycases", drawPoints = TRUE,
pointShape = "square", color = "darkblue" , label = "Moving Avg 7") %>%
dyRangeSelector(height = 20)%>%
dyLegend(width = 300)}
r1<-function(id){
dygraph(ts_peru_int_recovereddaily,
main = " Recovered Covid 19 Daily cases") %>%
dySeries("recovered_dailycases", stepPlot = TRUE,
fillGraph = TRUE, color = "turquoise" , label = "Recovered Daily Cases") %>%
dySeries("recovered7_dailycases", drawPoints = TRUE,
pointShape = "circle", color = "green" , label = "Moving Avg 7")%>%
dyRangeSelector(height = 20)%>%
dyLegend(width = 300)}
d1<-function(id){dygraph(ts_peru_int_deathsdaily,
main = "Covid 19 Daily Deaths") %>%
dySeries("deaths_dailycases", stepPlot = TRUE, fillGraph = TRUE,
color = "orange" , label = "Deaths Daily Cases") %>%
dySeries("deaths7_dailycases", drawPoints = TRUE, pointShape = "square",
color = "red" , label = "Moving Avg 7")%>%
dyRangeSelector(height = 20)%>%
dyLegend(width = 285)}Lastly, we use combineWidget to arrange our charts in the final graph. Additionally, we create the function write_alt_text to add an alternative text to our graph.
knitr::opts_chunk$set(echo = TRUE, fig.align="left")
write_alt_text <- function(
chart_type,
type_of_data,
reason,
source
){
glue::glue(
"{chart_type} of {type_of_data} where {reason}.<br> \n\nData source from {source}"
)
}
combineWidgets(
ncol = 2, colsize = c(2,1),
cumulates_plotly(1),
title = "Covid 19 Peru Interactive Time Series",
footer = write_alt_text(
"Time Series",
"confirmed, recovered and deaths cases from Covid 19 in Peru",
"informaction about the evolution of Covid 19 is needed",
"MINSA-Peru/ CSSE-John Hopkins University.<br>Made by Marco Arellano B.") ,
combineWidgets(
ncol = 1,
c1(2),
r1(3),
d1(4)))