In vjcitn/sars2app: COVID-19 data resources and analysis tools cut down for an interactive app

Introduction

suppressPackageStartupMessages({
library(sars2app)
})

The Arima* functions of sars2app model COVID-19 incidence series. Using ARIMA(0,1,3) by default we get an estimate of drift and the standard error of this estimate.

For the US as a whole, as of May 20 2020:

library(sars2app)
ej = enriched_jhu_data()
usa_full = Arima_nation(ej, max_date="2020-05-20")
usa_full

For the US dropping NY:

nyd = nytimes_state_data()
drny = Arima_drop_state(ej, nyd, max_date="2020-05-20")
drny

```r
par(mfrow=c(1,2), mar=c(4,3,2,2))
plot(usa_full, main="USA aggregated", ylim=c(15900,36000))
plot(drny, main="USA excluding NY", ylim=c(15900,36000) )

Two different data sources are used for this display, as certain anomalies were apparent on May 17 2020 that inhibited summing NYT data to get US aggregate. The reconciliation of NYT, JHU and USA facts daily data could be of interest.

Incidence velocity estimates for contiguous 48 states and DC

allst = contig_states_dc()
allarima = lapply(allst, function(x) Arima_by_state(nyd, x))
names(allarima) = allst
drifts = sapply(allarima, function(x) coef(x$fit)["drift"])
searima = function(a) sqrt(a$fit$var.coef["drift", "drift"])
se.drifts = sapply(allarima, searima)

library(rmeta)
meta.summaries(drifts, se.drifts)
```r
o = order(drifts)
metaplot(drifts[o], se.drifts[o], labels=allst[o], cex=.7, 
  xlab="Incidence velocity (CHANGE in number of confirmed cases/day)", ylab="State")
segments(rep(-350,46), seq(-49,-4), rep(-50,46), seq(-49,-4), lty=3, col="gray")

Optimization of AR and MA parameters

The dependence structure of fixed-length incidence series that are the focus of this package varies over time and between states. The min_bic function will choose BIC-minimizing AR and MA orders by searching over a grid. We have tools to survey states in this way from a specified time point, and to report summaries.

data(min_bic_2020_05_20)
min_bic_2020_05_20
min_bic_2020_05_20[["New York"]]

vjcitn/sars2app documentation built on Jan. 3, 2022, 12:19 a.m.