Package 'TACforecasting'

Generate forecasts of attendant care hours

Description

Generate forecasts from ETS and ARIMA models, reconcile them, and combine them. Return a fable object containing the forecasts.

Usage

get_forecasts(data, h, nsim)

Arguments

data	Data set computed from read_tac_data
h	Forecast horizon.
nsim	Number of simulated future sample paths per model.

---

Synthetic data for attendant hours by age group and injury group

Description

Artificial Transport Accident Commission attendant care data

Format

Time series of class 'tsibble'

Details

group_costs is a daily 'tsibble' with index 'billing_period' and two values:

adjusted_hours:	Total attendant care hours
nclaims:	Number of active claims

The data is disaggregated using two keys:

age_group:	Age group of client at the time of accident
injury_group:	Injury sustained by client due to accident

Source

Synthetic data

Examples

group_costs

---

Plot forecasts of attendant care hours disaggregated by age or injury group.

Description

Produce a time plot of attendant care hours per billing period for specific disaggregations.

Usage

plot_forecasts(
  forecasts,
  data,
  show_age_group = "<aggregated>",
  show_injury_group = "<aggregated>"
)

Arguments

forecasts	A fable object created by get_forecasts
data	The data used to construct the forecasts. This should be a tsibble object of the same form as group_costs.
show_age_group	A character string specifying either a specific age group or "<aggregated>" meaning the total across all age groups.
show_injury_group	A character string specifying either a specific injury group or "<aggregated>" meaning the total across all injury groups.

Author(s)

Rob J Hyndman

Examples

## Not run: 
  group_costs |> 
    get_forecasts(h=13, nsim=100) |> 
    plot_forecasts(group_costs)

## End(Not run)

---

Plot attendant care hours disaggregated by age or injury group.

Description

Produce a time plot of attendant care hours per billing period disaggregated by the given 'variable'

Usage

plot_total_hours(group_costs, variable = NULL, include_average = TRUE)

Arguments

group_costs	A tsibble containing costs optionally split by a variable
variable	Name of disaggregation variable. If NULL, aggregated costs are shown
include_average	Should the average cost per billing period be shown?

Author(s)

Rob J Hyndman

Examples

group_costs |> 
    plot_total_hours(age_group)

---

Read in TAC data

Description

This function takes two csv files as inputs: one containing the claims header and the other containing the attendant hours. It returns total hours per age group and injury group by billing period.

Usage

read_tac_data(claims_file, costs_file)

Arguments

claims_file	CSV file containing claims header
costs_file	CSV file containing attendant hours

Value

A tsibble object containing total attendant care adjusted hours for each billing period, disaggregated by age group and injury group. The column 'nclaims' shows the number of "active" claims in each billing period.

Author(s)

Rob J Hyndman

Examples

## Not run: 
group_costs <- read_tac_data(
  claims_file = "T086_claim_header.csv",
  costs_file = "T086_attendant_care_hours.csv"
)

## End(Not run)

---

Compute accuracy statistics

Description

Compute accuracy statistics

Usage

tac_accuracy(forecasts, actuals)

Arguments

forecasts	A fable object with forecasts, usually the output from get_forecasts
actuals	A tsibble with actual values. For example, the output from read_tac_data

Value

A tibble with accuracy statistics

---

Compute forecasts with a rolling origin and return accuracy statistics

Description

Compute forecasts with a rolling origin and return accuracy statistics

Usage

tscv_accuracy(group_costs, h, nsim, init, step)

Arguments

group_costs	A tsibble with actual values. For example, the output from read_tac_data
h	The forecast horizon
nsim	The number of simulations used in each forecast for each model.
init	The number of initial observations to use for the first fold.
step	The number of observations to skip between each fold.

Value

A tibble with accuracy statistics.

---