Apply sequential ARIMA pipeline

Applies a pipeline of (1) differencing, (2) autoregressive (AR) modeling, (3) moving average smoothing, and (4) band-pass filtering.

Usage

seqarima(
  x,
  d = NULL,
  p = NULL,
  q = NULL,
  fl = NULL,
  fu = NULL,
  ar.aic = "AIC",
  ar.collector = "median",
  ma.collector = "median",
  ar.args = list(),
  ma.args = list(),
  bp.args = list(),
  return.step = FALSE,
  verbose = TRUE
)

Arguments

x

A time series (`ts`) object.

d

Differencing order. Use 0 for no differencing, or 'auto' for KPSS-based auto-differencing.

p

AR order. Can be a single integer or a vector for ensemble.

q

MA order. Can be a single integer or a vector for ensemble.

fl

Lower frequency bound for band-pass filter.

fu

Upper frequency bound for band-pass filter.

ar.aic

Information criterion for AR model selection. One of 'AIC', 'BIC', 'FPE', 'AICc', 'KIC', 'AKICc'.

ar.collector

Collector for AR ensemble. One of 'mean', 'median', or 'pca'.

ma.collector

Collector for MA ensemble. One of 'mean', 'median', or 'pca'.

ar.args

A named list of additional arguments to be passed to the `Autoregressive()` function. This allows customization of the AR stage beyond standard parameters used in `burgar()`, etc. These arguments are merged with the core inputs via `do.call()`.

ma.args

A named list of additional arguments to be passed to the `MovingAverage()` function. This enables optional customization of the MA stage, for example by supplying `w.func`, custom weights, or smoothing parameters. Like `ar.args`, these are passed using `do.call()`.

bp.args

A named list of additional arguments to be passed to the `BandPass()` function. This enables optional customization of the MA stage, for example by supplying `resp` and `filt_order`, custom filter configurations.

return.step

Logical. If TRUE, intermediate stages are returned.

verbose

Logical. If TRUE, messages are printed during execution.

Value

A `ts` object with processed output and meta attributes.

Examples

if (FALSE) { # \dontrun{
# Generate noisy sinusoid
set.seed(123)
fs <- 1024
t <- seq(0, 1, by = 1 / fs)
x <- sin(2 * pi * 60 * t) + rnorm(length(t), sd = 0.5)
ts_obj <- ts(x, start = t[1], frequency = fs)

# Apply sequential ARIMA pipeline
out <- seqarima(
    ts_obj,
    d = 1,
    p = c(50, 100, 150),
    q = 20,
    fl = 30,
    fu = 300,
    ar.collector = "median",
    ma.collector = "mean",
    return.step = TRUE,
    verbose = FALSE
)

# Plot result
plot(out, main = "Output of seqarima()")

# View AR model metadata
attr(out, "meta")$ar_feat
} # }