ABSTRACT
Although mechanistic and phenomenological models are two of the main types of models in epidemic modeling, in COVID-19 studies, an increasing number of hybrid methods have been developed. This chapter focuses on ensembling various models to make hybrid models. It introduces two R packages: “forecastHybrid” and “opera” for forecast combinations, and dives into specific examples to illustrate the usage. The chapter considers the daily new death count for Florida. It illustrates the usage of opera using the daily new deaths time series for the state of Florida. Forecasting is of great importance in time series modeling. The chapter uses the forecast() function to produce a forecast for the next 14 days using the ensembling method. The prediction intervals are preserved from the individual component models. The R Package “opera” (online prediction by expert aggregation) provides methods to perform robust online predictions, for regression-oriented time series, by combining a set of user-provided forecasts.
