In rplzzz/vacovdata: COVID-19 Data for the Commonwealth of Virginia

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

library(vacovdata)
library(dplyr)
library(ggplot2)
library(ggthemes)
library(latex2exp)

Raw prevalence observations

These are the raw positive test fractions, statewide, uncorrected for enrichment due to targeted testing (what we call $\hat{f}$ in our paper).

pltdata <- filter(vacovdata, date > as.Date('2020-03-15'), !is.na(fpos)) %>%
  select(date, fpos, npos=positiveIncrease, nneg=negativeIncrease) %>%
  mutate(flo = qbeta(0.025, npos+1, nneg+1), fhi = qbeta(0.975, npos+1, nneg+1))

ggplot(pltdata, aes(x=date)) +
  geom_point(aes(y=fpos), size=1.5) +
  geom_errorbar(aes(ymin=flo, ymax=fhi)) +
  ylab(TeX('$\\hat{f}$')) +
  theme_bw()

Note, however, that there are several days in May that saw dramatic increases in the number of tests performed.

pltdata <- mutate(pltdata, ntest = npos+nneg)
## Will use these May correlations below
maydata <- filter(pltdata, date >= as.Date('2020-05-01'))
maycor <- signif(cor(maydata$ntest, maydata$fpos, method='spearman'),2)

ggplot(pltdata, aes(x=date, y=ntest)) +
  geom_line(size=1.2) +
  ylab('Number of tests') +
  theme_bw()

These excess tests, relative to the surrounding days, tend to be negative, producing low outlier points in the plot of $\hat{f}$, particularly on May 1st, 15th, 20th, and 24th. Overall for the month of May, the correlation between the number of tests performed and the estimate of $\hat{f}$ (measured by the Spearman correlation coefficient, $\rho_s$) is $\rho_s = r maycor$. This high correlation suggests that on the days with a surge in testing, a different selection procedure is being used, resulting in a different enrichment factor.

Growth rates

Both the positive test fraction and the growth rates derived from it are quite noisy, so we use a 7-day sliding window to accumulate observations.

grpltdata <- select(vacovdata, date, pos=positive, tot=totalTestResults) %>%
  fitgr(7) %>% 
  filter(date > as.Date('2020-03-23'), !is.na(kobs))

ggplot(grpltdata, aes(x=date)) +
  geom_point(aes(y=kobs, color='uncorrected'), size=1.5) +
  geom_smooth(aes(y=kobs, color='uncorrected'), se=FALSE) +
  geom_point(aes(y=kpop, color='corrected'), size=1.5) +
  geom_smooth(aes(y=kpop, color='corrected'), se=FALSE) +
  #coord_cartesian(ylim=c(-0.1,0.3)) +
  #xlim(as.Date(c('2020-04-01', '2020-05-30'))) +
  scale_color_solarized() +
  theme_bw()