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

Description Usage Arguments Value Examples

Identify change point using a loss function method and find expected SSD visits/calculate misses. This is done by fitting a linear model for data prior to i, with i ranging from period 0 to max period - 2. For a specified loss function, the first local minima is used used to identifiy the optimal change point

find_cp_loss_fun(
  data,
  var_name = "n_miss_visits",
  return_miss_only = FALSE,
  return_loss_fun_tab_only = FALSE,
  week_period = FALSE,
  specify_cp = NULL,
  loss_function = "RMSE"
)

`data`	A dataframe output by count_prior_events_truven
`var_name`	A character string of outcome for which to apply analysis
`return_miss_only`	Logical to only return miss information
`return_loss_fun_tab_only`	Logical to only return loss function table that includes estimates for all loss functions available
`week_period`	Logical to incorporate a "day of the week" effect into the linear model. Note this is only sensible for one-day period aggregation.
`specify_cp`	Set a specific change point you want to use instead of searching for optimal change point. Enter a postive integer value repersenting the days before the index on which you you want to specify the change point. (e.g. 100 would be 100 days before the index)
`loss_function`	The loss function used to identify the change point (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)

A list containing miss information, changepoint information, predictions, the model itself, and a plot of the middle finger curve and model.

cp_result_pettit <- final_time_map %>%
filter(days_since_dx >= -180) %>%
count_prior_events_truven(event_name = "any_ssd", start_day = 1, by_days = 1) %>%
find_cp_loss_fun(var_name = "n_miss_visits", return_miss_only = FALSE, week_period=TRUE, loss_function = "MAE")