Forecasting epidemic trajectories: Time Series Growth Curves package tsgc
Introduction | Theory | Gompertz curve | Dynamic Gompertz model | State space form and estimation | Forecasts and peak prediction | Reproduction Number $R_t$ | Reinitialization | Reinitializing the data series | Reinitializing the model | Functionality of tsgc | Leading indicator model | Forecast accuracy and rolling-origin validation | Illustration of the tsgc package | Setting up the forecasting exercise | Estimation | Model diagnostics | Results | Triggering reinitialization | Estimating the reinitialized model | Leading indicator model: England cases and hospitalisations | Adding regressors and future covariate paths | Comparing the dynamic Gompertz and leading-indicator models | Selecting the lag by rolling cross-validation | Extensions: non-epidemiological applications and alternative data frequencies | Conclusions | Acknowledgements | References | Appendix: Incorporating seasonal terms into the state space model | Appendix: How to import data from CSV file | Appendix: Controlling the plot x-axis with idx_axis_opts() | mode = "date" (the default here, since gauteng_cal$posixct = TRUE) | mode = "position" | mode = "steps" | mode = "time_since" | Adding an info box | Appendix: Compound calendar steps with idx_step() and idx_calendar_step() | Appendix: Heterogeneous repeating cycles with multi_step_pattern() and idx_calendar_multi_step() | Appendix: Non-calendar idx_calendar anchors and idx_offset_to_pos()
Rendered fromtsgc_vignette.Rmdusingknitr::rmarkdown