compare_cp_methods_bootstrap: Run multiple bootstrapped change_point comparisons (in...

In aarmiller/delayDX: Tasks for findind and estimating diagnostic delays

Description

Run multiple bootstrapped change_point comparisons (in parallel)

Usage

compare_cp_methods_bootstrap(
  sim_data,
  boot_trials = 100L,
  num_cores = NULL,
  method1,
  method2,
  loss_fun
)

Arguments

sim_data	A dataset containing the time_map, miss bins and other parameters used for the simulation. This dataset should be created using the 'prep_sim_data()' function
boot_trials	The number of bootstrapped trials to run (default is 100)
num_cores	The number of worker cores to use. If not specified will determined the number of cores based on the which ever is the smallest value between number of boot_trials or detected number of cores - 1
method1	First change point method to compare
method2	Second change point method to compare
loss_fun	The loss function used in the comparison (mean squared error (MSE), root mean squared error (RMSE), mean absolute error (MAE), mean squared log error (MSLE), or root mean squared log error (RMSLE)
week_period	Logical to incorporate a "day of the week" effect into.
eval_criteria	The evaluation criteria used to find change points, if using a linear regression method

Value

A tibble with two columns: 1) the bootstrap trial, 3) results from each trial (each trial contains a nested tibble with 3 columns: 1) the loss function used, 2) the value for method 1, and 3) the value for method 2)

Examples

#load example final_time_map dataset
load("/Shared/Statepi_Diagnosis/grant_projects/hsv_enceph/scripts/validation/enrolled_ge_365/report_data.RData")

# rename ED column
final_time_map <- final_time_map %>% rename(ed = ED)

# run prep sim function (note = enter any cp_method here, it just acts as a place holder)
tmp_sim_data <- prep_sim_data(final_time_map, event_name = "any_ssd", cp_method = "lm", start_day = 1L, by_days = 1L,
                              week_period = TRUE)

# run bootstrapped comparisons between two method (e.g. lm_cube vs. cusum) based on MSE
results <- compare_cp_methods_bootstrap(sim_data = tmp_sim_data,
                                        boot_trials = 1000L,
                                        num_cores = NULL,
                                        method1 = "lm_cube",
                                        method2 = "cusum",
                                        loss_fun = "MSE")

aarmiller/delayDX documentation built on July 11, 2021, 4:01 p.m.