get_submission_via_trajectory_simulation: Get log scores and full predictive distributions for each...

In reichlab/2017-2018-cdc-flu-contest: code for cdc flu contest 2017-2018

Description

for a given season using a predictive method that works by simulating trajectories of incidence in each remaining week of the season. Results are stored in a data frame, saved in a .rds file with a name like "model_name-region-season-loso-predictions.rds" Results have columns indicating the analysis time season and season week, model name, log scores for each prediction target, the "log score" used in the competition (adding probabilities from adjacent bins) for each prediction target, as well as the log of the probability assigned to each bin.

Usage

get_submission_via_trajectory_simulation(data, analysis_time_season,
  first_analysis_time_season_week = 10, last_analysis_time_season_week = 41,
  prediction_target_var, incidence_bins, incidence_bin_names, n_trajectory_sims,
  simulate_trajectories_function, simulate_trajectories_params,
  all_regions = c("National", paste0("Region ", 1:10)))

Arguments

data	data frame with observed disease data available to use for predictions
analysis_time_season	character vector of length 1 specifying the season to obtain predictions for, in the format "2000/2001"
first_analysis_time_season_week	integer specifying the first week of the season in which to make predictions, using all data up to and including that week to make predictions for each following week in the season
last_analysis_time_season_week	integer specifying the last week of the season in which to make predictions, using all data up to and including that week to make predictions for each following week in the season
prediction_target_var	string specifying the name of the variable in data for which we want to make predictions
incidence_bins	a data frame with variables lower and upper defining lower and upper endpoints to use in binning incidence
incidence_bin_names	a character vector with a name for each incidence bin
n_trajectory_sims	integer number of trajectories to simulate
simulate_trajectories_function	a function to call to simulate incidence trajectories. It will be called with the following arguments: * n_sims = number of trajectories to simulate * max_prediction_horizon = number of following weeks to simulate * data = all available data to use in doing simulation, up to and including the analysis time * region = region * analysis_time_season = analysis_time_season * analysis_time_season_week = week of the season at which we are making the predictions * params = simulate_trajectories_params; additional user-provided parameters
simulate_trajectories_params	optional additional parameters to pass to simulate_trajectories_function
all_regions	a character vector with all of the regions over which to make forecasts, defaults to HHS region names

Value

data frame with contents suitable for writing out as a csv file in the CDC's standardized format

reichlab/2017-2018-cdc-flu-contest documentation built on May 24, 2019, 6:17 a.m.