Dear members,

Since a few years, I have been using R to do projections and predictive analysis using R for COVID-19 epidemiological situation in Kosovo. I’ve been using EpiNow2 package, with other personal coding as well. This package estimates the time-varying reproduction number, growth rate, and doubling time using a range of open-source tools (Abbott et al.), and current best practices (Gostic et al.). It aims to help users avoid some of the limitations of naive implementations in a framework that is informed by community feedback and is actively supported.

The coding is below:

# Aktivizimi i pakos kryesore
library(EpiNow2)

# nese nuk eshte e instaluar, duhet te perdorni komanden install.packages(EpiNow2) dhe pas instalimit, ta aktivizoni me komanden e mesiperme

# percaktimi i numrit te procesoreve te kompjuterit qe nevojiten me i pergatit modelimet
options(mc.cores = 4)

# referenca nga literatura – parametrat nderkomtare te SARS-CoV-2 (koha e inkubacionit, generacionit)
generation_time <- get_generation_time(disease = “SARS-CoV-2”, source = “ganyani”, max=10)
incubation_period <- get_incubation_period(disease = “SARS-CoV-2”, source = “lauer”, max=10)

# definimi i vonesave ne raportim si logaritem i mesatares prej 4 ditesh dhe sd prej 1 ditesh ne mungese te evidences.
reporting_delay <- list(mean = convert_to_logmean(4, 1),
mean_sd = 0.1,
sd = convert_to_logsd(4, 1),
sd_sd = 0.1,
max = 15)

# periudha e hulumtimit te rasteve per modelim
dates <- seq(as.Date(“2023-07-01”), as.Date(“2023-08-31”), by = “days”)

# numri i rasteve per secilen nga datat e mesiperme (bashkengjitur e ke edhe nje Excel sheet)
confirm <- c(0,0,0,0,2,0,0,0,0,0,0,0,0,0,0,2,1,1,1,0,0,0,0,1,0,0,1,1,0,0,0,0,1,0,0,0,0,0,0,1,1,1,0,1,2,4,3,7,6,5,3,12,0,4,5,8,2,1,6,4,6,7)

# krijimi i dataset per rastet e konfirmuara e te raportuara gjate diteve paraprake
reported_cases <- data.frame(date = dates, confirm = confirm)

# PERSHKRIMI I ASAJ SE CFARE BEN PAKOJA EPINOW2
# estimate Rt and nowcast/forecast cases by date of infection
# on a 4 core computer this example should take between 2 ~ 5 minutes to run
# to see saved logs view the dated logs folder
# to see saved results see the dated results folder
# some data sets may produce divergent transition warnings
# this is not unexpected and is usually fine (if dts < 1% of samples)
# but is an area of research as we seek to optimise the underlying model.
# If you have some prior knowledge of the scaling between observations and
# reports see ?obs_opts for options.
# If you have some prior knowledge on the truncation in your data or multiple
# snapshots of data see ?trunc_opts for options.
# Note that the default settings may not be appropriate for your use case.
# Kodi eshte i pergatitur perfundimisht. Kujdes me ndryshimet e meposhtme.
out <- epinow(reported_cases = reported_cases,
generation_time = generation_time,
delays = delay_opts(incubation_period, reporting_delay),
rt = rt_opts(prior = list(mean = 1.5, sd = 0.5)),
# here we define the quality of the gaussian process approximation
# if the fit to data appears poor try increasing basis_prop and
# potentially the boundary_scale (see ?gp_opts for details)
# though this will likely increase runtimes.
gp = gp_opts(basis_prop = 0.2),
# in some instances stan chains can get stuck when estimating in
# these instances consider using the future fitting mode by passing
# `future = TRUE, max_execution_time = 60 * 30` to stan_opts and calling
# `future::plan(“multiprocess”)` prior to running epinow this will time out
# chains after 30 minutes but still return results from completed chains
stan = stan_opts(),
horizon = 14,
target_folder = “results”,
logs = file.path(“logs”, Sys.Date()),
return_output = TRUE,
verbose = TRUE)

# krijimi i tabeles per te dhenat e pergjithshme
summary(out)

# paraqitja grafike e projeksioneve
plot(out)

# krijimi i tabeles me knit
knitr::kable(summary(out))

If any support is needed, hit me up! I can also build customized R code for your analysis and projections as well.