Arima_drop_state: fit ARIMA model to US data dropping one state

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

Description

fit ARIMA model to US data dropping one state

Usage

Arima_drop_state(
  src_us,
  src_st,
  state.in = "New York",
  MAorder = 2,
  Difforder = 1,
  basedate = "2020-02-15",
  lookback_days = 29,
  ARorder = 0,
  max_date = NULL
)

Arguments

src_us	tibble for national level data like that of enriched_jhu_data()
src_st	tibble for state level data like that of nytimes_state_data()
state.in	character(1) state name
MAorder	numeric(1) order of moving average component
Difforder	numeric(1) differencing order d of ARIMA(p,d,q)
basedate	character(1) used by lubridate::as_date to filter away all earlier records
lookback_days	numeric(1) only uses this many days from most recent in src
ARorder	order of autoregressive component
max_date	a date from which to start lookback ... defaults to NULL in which case the latest available date is used

Value

instance of S3 class Arima_sars2pack

Note

Apparent discrepancies in counts in vicinity of April 15 between NYT and JHU for full US incidence lead us to employ the two sources for the example. If NYT alone is used, summing to obtain USA overall, the resulting USA incidence trend is extravagantly variable.

Examples

ej = enriched_jhu_data()
usa_full = Arima_nation(ej)
plot(usa_full)
nyd = nytimes_state_data()
drny = Arima_drop_state(ej, nyd)
drny
plot(drny)
opar = par(no.readonly=TRUE)
par(mar=c(4,3,2,2), mfrow=c(1,2))
plot(usa_full, main="all states", ylim=c(17000,38000))
plot(drny, main="excluding NY", ylim=c(17000,38000))
par(opar)

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