rlgt function

Fit an Rlgt model

The main function to fit an rlgt model. It fits the parameter values with MCMC.

rlgt(
  y,
  seasonality = 1,
  seasonality2 = 1,
  seasonality.type = c("multiplicative", "generalized"),
  error.size.method = c("std", "innov"),
  level.method = c("HW", "seasAvg", "HW_sAvg"),
  xreg = NULL,
  control = rlgt.control(),
  verbose = FALSE,
  method = "Stan",
  experimental = "",
  homoscedastic = F
)

Arguments

• y: time-series data for training (provided as a numeric vector, or a ts, or msts object).
• seasonality: This specification of seasonality will be overridden by frequency of y, if y is of ts or msts class. 1 by default, i.e. no seasonality.
• seasonality2: Second seasonality. If larger than 1, a dual seasonality model will be used. However, this is experimental. If not specified and multiple seasonality time series (of msts class) is used, a single seasonality model will be applied, one with seasonality equal to the largest of seasonalities of the time series. 1 by default, i.e. no seasonality or single seasonality.
• seasonality.type: Either "multiplicative" (default) or "generalized". The latter seasonality generalizes additive and multiplicative seasonality types.
• error.size.method: Function providing size of the error. Either "std" (monotonically, but slower than proportionally, growing with the series values) or "innov" (proportional to a smoothed abs size of innovations, i.e. surprises)
• level.method: "HW", "seasAvg", "HW_sAvg". Here, "HW" follows Holt-Winters approach. "seasAvg" calculates level as a smoothed average of the last seasonality number of points (or seasonality2 of them for the dual seasonality model), and HW_sAvg is an weighted average of HW and seasAvg methods.
• xreg: Optionally, a vector or matrix of external regressors, which must have the same number of rows as y.
• control: list of control parameters, e.g. hyperparameter values for the model's prior distributions, number of fitting interations etc.
• verbose: whether verbose information should be printed (Boolean value only), default FALSE.
• method: Sampling method, default Stan.
• experimental: Run different versions ("nostudent", "noglobal", "nohet", "ets") for ablation studies
• homoscedastic: Run with homoscedastic or heteroscedastic version of the Gibbs sampler version. By default it is set to FALSE, i.e., run a heteroscedastic model.

Returns

rlgtfit object

Examples

# The following is a toy example that runs within a few seconds. To get good 
# fitting results the number of iterations should be set to at least 2000, and 
# 4 chains should be used (the default). To speed up computation the number of 
# cores should also be adjusted (default is 4).

rlgt_model <- rlgt(lynx, 
       control=rlgt.control(MAX_NUM_OF_REPEATS=1, NUM_OF_ITER=50, NUM_OF_CHAINS = 1, 
                            NUM_OF_CORES = 1), verbose=TRUE)

# print the model details
print(rlgt_model)

## Not run: demo(exampleScript)