yardstick (version 0.0.4)

mase: Mean absolute scaled error

Description

Calculate the mean absolute scaled error. This metric is scale independent and symmetric. It is generally used for comparing forecast error in time series settings. Due to the time series nature of this metric, it is neccesary to order observations in ascending order by time.

Usage

mase(data, ...)

# S3 method for data.frame mase(data, truth, estimate, m = 1L, mae_train = NULL, na_rm = TRUE, ...)

mase_vec(truth, estimate, m = 1L, mae_train = NULL, na_rm = TRUE, ...)

Arguments

data

A data.frame containing the truth and estimate columns.

...

Not currently used.

truth

The column identifier for the true results (that is numeric). This should be an unquoted column name although this argument is passed by expression and supports quasiquotation (you can unquote column names). For _vec() functions, a numeric vector.

estimate

The column identifier for the predicted results (that is also numeric). As with truth this can be specified different ways but the primary method is to use an unquoted variable name. For _vec() functions, a numeric vector.

m

An integer value of the number of lags used to calculate the in-sample seasonal naive error. The default is used for non-seasonal time series. If each observation was at the daily level and the data showed weekly seasonality, then m = 7L would be a reasonable choice for a 7-day seasonal naive calculation.

mae_train

A numeric value which allows the user to provide the in-sample seasonal naive mean absolute error. If this value is not provided, then the out-of-sample seasonal naive mean absolute error will be calculated from truth and will be used instead.

na_rm

A logical value indicating whether NA values should be stripped before the computation proceeds.

Value

A tibble with columns .metric, .estimator, and .estimate and 1 row of values.

For grouped data frames, the number of rows returned will be the same as the number of groups.

For mase_vec(), a single numeric value (or NA).

Details

mase() is different from most numeric metrics. The original implementation of mase() calls for using the in-sample naive mean absolute error to compute scaled errors with. It uses this instead of the out-of-sample error because there is a chance that the out-of-sample error cannot be computed when forecasting a very short horizon (i.e. the out of sample size is only 1 or 2). However, yardstick only knows about the out-of-sample truth and estimate values. Because of this, the out-of-sample error is used in the computation by default. If the in-sample naive mean absolute error is required and known, it can be passed through in the mae_train argument and it will be used instead. If the in-sample data is available, the naive mean absolute error can easily be computed with mae(data, truth, lagged_truth).

References

Rob J. Hyndman (2006). ANOTHER LOOK AT FORECAST-ACCURACY METRICS FOR INTERMITTENT DEMAND. Foresight, 4, 46.

See Also

Other numeric metrics: ccc, huber_loss_pseudo, huber_loss, iic, mae, mape, rmse, rpd, rpiq, rsq_trad, rsq, smape

Other accuracy metrics: ccc, huber_loss_pseudo, huber_loss, iic, mae, mape, rmse, smape

Examples

Run this code
# NOT RUN {
# Supply truth and predictions as bare column names
mase(solubility_test, solubility, prediction)

library(dplyr)

set.seed(1234)
size <- 100
times <- 10

# create 10 resamples
solubility_resampled <- bind_rows(
  replicate(
    n = times,
    expr = sample_n(solubility_test, size, replace = TRUE),
    simplify = FALSE
  ),
  .id = "resample"
)

# Compute the metric by group
metric_results <- solubility_resampled %>%
  group_by(resample) %>%
  mase(solubility, prediction)

metric_results

# Resampled mean estimate
metric_results %>%
  summarise(avg_estimate = mean(.estimate))
# }

Run the code above in your browser using DataCamp Workspace