speclagreg: Estimate regresson operators in a lagged linear model

In freqdom: Frequency Domain Analysis for Multivariate Time Series

Description

Estimate regresson operators in a lagged linear model using spectral methods. Assume model

Y_t = ∑_{k=-q}^p A_k X_{t-k} + \varepsilon_t

where X_t is a stationary multivariate time series, (A_k)_{-q ≤q k ≤q p} is a filter and \varepsilon_t is white noise. Function speclagreg estimates parameters A_k with k \in lags

Usage

speclagreg(X, Y, Kconst = 1, K = NULL, lags = 0:0, freq = NULL,
  p = 10, q = 10, weights = "Bartlett")

Arguments

X	first process
Y	second process, if null then autocovariance of X is computed
Kconst	used for heuristic as in reg.dim.est
K	dimension for inversion if no heuristic should be used
lags	which A_k should be estimated
freq	grid of frequencies for computation as in fourier.transform
p	window size for estimation of spectral density of X
q	window size for estimation of spectral density of Y and X
weights	as in spectral.density

Value

timedom operators

See Also

linproc

Examples

X = rar(100)
Y = rar(100)
#estimate regressors in model $Y_t = \sum_{i\in Z} A_i X_{t-i}$
A = speclagreg(X,Y)
# check an advanced examples in demo(lagged.reg)

freqdom documentation built on May 2, 2019, 5:55 p.m.