Articles • epinowcast

All vignettes

ARIMA latent residuals: maths, priors, and usage: The maths behind the ARIMA(p, d, q) latent residuals in epinowcast, how to use them in any module’s formula, and how to set their priors.
Estimating reporting delays with the full and delay-only models: A walk through of estimating a reporting delay distribution, comparing the full nowcasting model with the delay-only model that conditions on known totals.
Discretised distributions: Distributions and their discretisation in epinowcast
Getting Started with Epinowcast: Nowcasting: A quick start example demonstrating use of epinowcast to nowcast hospital admissions.
Model Features Summary: Quick reference to package capabilities
Gaussian process latent terms: maths, priors, and usage: The maths behind the Hilbert-space approximate Gaussian process latent terms in epinowcast, how to use them in any module’s formula, and how to set their priors.
Hierarchical nowcasting of age stratified COVID-19 hospitalisations in Germany: A case study exploring hierarchical models of varying complexity to jointly nowcast age stratified COVID-19 hospitalisations in Germany.
Comparing Inference Methods: A comparison of NUTS sampling, pathfinder approximate inference, and pathfinder-initialised NUTS across two model specifications.
Latent process and periodic options for the growth-rate model: Random walks, ARIMA(p, d, q) residuals, and periodic effects — the time-series structures available in the formula interface for the growth rate.
Model definition and implementation: Model formulation and implementation details
Case studies: A place to document how epinowcast has been used
Visualising Preprocessed Data: Understanding reporting patterns before model fitting.
Estimating the effective reproduction number in real-time for a single timeseries with reporting delays: A walk through of a simple approach to jointly estimating the effective reproduction number over time and the delay from a positive test to this test being reported.
Resources to help with model fitting using Stan: How to address issues you may encounter with Stan
Temporal aggregation guide: How to fit nowcasts when the data and process timesteps differ, including pure weekly, weekly reporting on a daily process, and a daily benchmark.