March 7, 2020

Overview

  • A curiousity timeline
  • Some COVID-19 information
  • A little R and some visualisations
  • Some modelling and estimation

Timeline

  • 22nd Jan: Whatsapp
    • Confirmed cases: 555
  • 25th Jan: Whatsapp
    • Confirmed cases: 1,434

John Hopkins Live Tracker Dashboard

Timeline

  • 22nd Jan: Whatsapp
    • Confirmed cases: 555
  • 25th Jan: Whatsapp
    • Confirmed cases: 1,434
  • 10th Feb: SatRday talk submission
    • Confirmed cases: 42,000

Timeline

  • 22nd Jan: Whatsapp
    • Confirmed cases: 555
  • 25th Jan: Whatsapp
    • Confirmed cases: 1,434
  • 10th Feb: SatRday talk submission
    • Confirmed cases: 42,000
  • 13th Feb: Prof. Vardas @NRF
    • Confirmed cases: 60,429

COVID-19

\(R_0\): The number of people, on average, that one sick person will infect. cboettig gist

COVID-19

COVID-19 Origins

COVID-19 Clinical Assessment (T8 -> T11)

Coronaviruses

6 already known to infect humans:

  • hCoV-OC63
  • hCoV-NL63
  • hCoV-HKU1

  • hCoV-229E

  • SARS-CoV 2002/2003 Guangdong, China

  • MERS-CoV 2012 Middle East (ongoing)

SARS

Easy numbers to remember:

  • Outbreak lasted 8 months
  • Confirmed cases: 8,000
  • Total deaths: 800

COVID-19

library(scales)
scale_y_continuous(label = comma)

COVID-19 Data

COVID-19 Latest

COVID-19 Latest (Mortality Rates)

Mortality Rates (China & !China)

FAIL 

ggplot() + scale_y_log10()

Corrected Mortality Rates (smooth infections)

smooth.spline(data$date, data$confirmed, spar = .5)

Corrected Mortality Rates (smooth deaths)

Corrected Mortality Rates

ggplot() + stat_smooth(method = 'lm')

Corrected Mortality Rates

Lag : 5 days from confirmed -> deceased

Mortality rate: 3.7%

COVID-19 In Mainland China

library(sf)
geom_sf(data = china_provinces)

Can we do better?

Express the totals on an axis.

Drop the geography. Rather split on countries.

Focus on Top-n elements.

Perhaps scale the totals?

Elsewhere

Outbreak status by local gradient (ggplot)

Outbreak status by local gradient (d3)

Shifting the x-axis to match the outbreak stage

Build a “typical” growth line.

Shift each trajectory by country to “best-fit” existing trends.

Modelling the typical growth (0:12)

Modelling the typical growth (0:24)

Putting it together

  • Mortality Rates
  • Quarantine Policies
  • Economic

R packages

  • ioslides : for markdown-based presentations
  • tidyr : for data preparation
  • dplyr : to target an output format
  • ggplot2 : for visualisation
  • r2d3 : for better visualisation