GUIDE.MD

In reichlab/cdcForecastUtils: Utility functions for CDC forecasts

To use package modelers need two functions

• Fit method: takes a time series of arbitrary length and fits the model to the time series
• Predict method: takes a fitted model and produces a sample matrix of nsim*h

Steps to use package

• Fit model to data
• Compute h using the utility time_left_in_season <- get_time_left_in_season(current_date_in_cdc_format,season_end)
• Produce a matrix of nsim*time_left_in_season using predict method called the trajectory matrix
• (Optional) Apply custom backfill model to observed data
• Compute time since start of season using time_from_start_of_season <- get_time_from_start_of_season(season_start,current_date_in_cdc_format)
• If backfill: bind together backfill matrix of nsim*time_from_start_of_season and trajectory matrix
• If no backfill: create a matrix of rep(observed_data,nsim) and cbind to trajectory matrix
• Call trajectories_to_short_term_and_seasonal_binned_distributions to produce the submission df
• Call generate_csv_from_submission_df to produce the csv

reichlab/cdcForecastUtils documentation built on May 6, 2020, 10:43 a.m.