Motivation

Because of the COVID-19 Pandemia, Argentine Federal Goverment impossed a total quarantine for the whole national territory on March 19 of 2020. Since Argentine’s Public Health ministry publish the statistics to the World Health Organization I’ve decided to exploit the information extracted from the WHO by JHU CSSE to monitor the Argentinian case and somehow measure the effect of the quarentine.

Argentine Quarentine Facts

• Initially defined to be in effect untill March 31 of 2020
• Enforced by law and semi-rigid controlled by all Argentinian law enforcement forces
• Moderate to high civic obedience

Shiny applicaton and datasets

The application hosted at: https://gdmuzzillo.shinyapps.io/COVID_19_Argentina is based on the following datasources prepared by JHU CSSE labs:

• Global Time Series of Confirmed Cases
• Global Time Series of Recovered Patients
• Global Time Series of Covid-19 caused Deaths

– Sumary of Confirmed cases

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##     0.0    18.0   128.0   455.9   782.5  1715.0

– Sumary of recovered patients

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##    0.00    0.00    3.00   82.69  150.00  358.00

Data preparation

• Data comming from CSV needs to be transformed into R real time series

    # Read CSV Datasets
    dfConfirmed <- read.csv(file = "time_series_covid19_confirmed_global.csv")
    dfDeaths <- read.csv(file= "time_series_covid19_deaths_global.csv")
    dfRecovered <- read.csv(file = "time_series_covid19_recovered_global.csv")
    # Filter for argentinian events
    dfConf <- dfConfirmed[dfConfirmed$Country.Region=="Argentina",]
    dfDth <- dfDeaths[dfDeaths$Country.Region=="Argentina",]
    dfRec <- dfRecovered[dfRecovered$Country.Region=="Argentina",]
    #Transform into a proper time series Data Frames
    tsConfirm <-  melt(dfConf,c("Province.State","Country.Region","Lat","Long"),na.rm = TRUE)
    tsDeath <-  melt(dfDth,c("Province.State","Country.Region","Lat","Long"),na.rm = TRUE)
    tsRecovered <-  melt(dfRec,c("Province.State","Country.Region","Lat","Long"),na.rm = TRUE)
    #Transform date columns (Variable) to be ISO compatible Dates
    tsConfirm$variable <- sub("X","",tsConfirm$variable)
    tsDeath$variable <- sub("X","",tsDeath$variable)
    tsRecovered$variable <- sub("X","",tsRecovered$variable)
    tsConfirm$variable <- as.Date(tsConfirm$variable, format = "%m.%d.%y")
    tsDeath$variable <- as.Date(tsDeath$variable, format = "%m.%d.%y")
    tsRecovered$variable <- as.Date(tsRecovered$variable, format = "%m.%d.%y")

    monthDate <- as.Date( paste( "2020", "03" , "01", sep = "-"), format = "%Y-%m-%d")
    
    tsConfirm <- tsConfirm[tsConfirm$variable >= as.Date(monthDate),]
    tsRecovered <- tsRecovered[tsRecovered$variable >= as.Date(monthDate),]
    tsDeath <- tsDeath[tsDeath$variable >= as.Date(monthDate),]
    tsPred <- xts(x = tsConfirm$value, order.by = tsConfirm$variable)
    tsPredDm <- dynlm(formula=tsConfirm$value ~ trend(tsConfirm$value)+L(tsConfirm$value), data = tsConfirm )
    tsPred1 <- predict(tsPredDm)
    tsConfirm$pred <- as.ts(tsPred1)
    tsConfirm$recovered <- tsRecovered$value
    tsConfirm$deaths <- tsDeath$value
    tsConfirm$diffConfirmed <- as.zoo(diff(tsPred,lag=1,differences=1, na.pad = TRUE))

Plot of Facts during the Quarantine vs Predicted Cases

• Plot of Time Series for predicted Cases by Quarentine Day

Plot of the Delta cases per Quarantine Day

• Plot of Time Series for predicted Cases by Quarentine Day