sample_predictive_trajectories_arima_wrapper: A wrapper around sample_predictive_trajectories_arima...

Description Usage Arguments Value

View source: R/sarima-utils.R

Description

This function does a few things worth noting. It subsets data to the analysis time. It pulls in an appropriate SARIMA fit, either a leave-one-season-out fit if the analysis time is before the first test season or a fit based on all of the training data if the analysis time is in or after the first test season. It linearly interpolates missing values. It handles log transformations and possible seasonal differencing that may be done outside of functionality in the forecast package.

Usage

1
2
sample_predictive_trajectories_arima_wrapper(n_sims, max_prediction_horizon,
  data, region, analysis_time_season, analysis_time_season_week, params)

Arguments

n_sims

number of trajectories to simulate

max_prediction_horizon

how many steps ahead to simulate

data

data set

region

region

analysis_time_season

season in which we're predicting

analysis_time_season_week

week of the season in which we're making our predictions, using all data up to the analysis time to make predictions for later time points

params

other parameters. A list with the following entries: * fits_filepath = path to a directory where SARIMA model fits are located * prediction_target_var = string naming variable in data we are predicting * seasonal_difference = logical specifying whether a seasonal difference should be computed manually before passing to auto.arima * transformation = string, either "log", "box-cox", or "none", indicating type of transformation to do * first_test_season = string, in format of "2011/2012", specifying first test season.

Value

an n_sims by h matrix with simulated values


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