TERROR Quality Control of Outliers

TRAMO for ERRORs (TERROR) controls the quality of the data by checking outliers at the end of the series

Usage

terror(
  ts,
  spec = c("trfull", "tr0", "tr1", "tr2", "tr3", "tr4", "tr5"),
  nback = 1,
  context = NULL
)

Arguments

ts

a univariate time series.

spec

the model specification. Can be either the name of a predefined specification or a user-defined specification.

nback

number of last observations considered for the quality check.

context

the dictionnary of variables.

Value

a mts object with 7 variables:

• actual the actual data at the end of the series.

• forecast the forecast of the actual data at the end of the series.

• error the absolute errors (= observed - forecasts).

• rel.error relative errors ("scores") : ratios between the forecast errors and the standard deviation of the forecasts of the last observations (positive values mean under-estimation).

• raw the transformed series. More especially, if the chosen model implies a log-transformation, the values are obtained after a log-transformation. Other transformations, such leap year corrections or length-of periods corrections may also be used.

• fraw the forecast of the transformed series.

• efraw the absolute errors of the transformed series.

Examples

terror(rjd3toolkit::ABS$X0.2.09.10.M, nback = 2)
#>          actual forecast     error rel. error transformed tr.fcast   tr.error
#> Jul 2017 1445.5 1468.903 -23.40336 -0.4328439    7.276211 7.292271 0.03710535
#> Aug 2017 1303.1 1341.213 -38.11282 -0.7664734    7.172501 7.201330 0.03761155