pTimEv: Model stationnary testing

Description

Estimate the parameters variations of a model of canonical form considering a sliding window on an external dataset.

Usage

pTimEv(
  TS,
  nVar,
  dMax,
  TSdate,
  whatTerms = NULL,
  weight = NULL,
  wlength = 1000,
  onestep = 100,
  removeExtr = 1
)

Value

A list containing:

$slidingoutGM An n*(pMax+1) matrix presenting the pMax estimated parameters p1(t), p2(t) etc. column by column. The residual signal epsilon(t) is provided in the last (i.e. pMax + 1) column. Each line correspond to one date provided in $TSdate

$TSdate A time vector relating to the estimates presented in $slidingoutGM

$W A vector providing the output values that can kept (=1) or must be removed (=0)

$whatTerms A vector recalling the terms taken into account in the analysis (their order refers to poLabs(nVar,dMax) function)

$param A vector with the parameter values used to apply the function: nVar, dMax, wlength, onestep, removeExtr

Arguments

TS: The time series to be tested
nVar: Number of variables considered in the polynomial formulation.
dMax: Maximum degree of the polynomial formulation.
TSdate: The time vector
whatTerms: The terms to be considered in the analysis. Note that these are organised following the convention defined by poLabs(nVar,dMax). Since only the structure is required, if coefficients are provided, these are transformed to 1.
weight: A vector providing the binary weighting function of the input data series (0 or 1). By default, all the values are set to 1.
wlength: The window length
onestep: Step length between two estimations
removeExtr: Ratio of estimated values to be removed (if chosen equal to 0.1, only 90 disersion will be kept)

Author

Sylvain Mangiarotti

Examples

Run this code

#Example
data(TS)
plot(TS[,1], TS[,2], type='l')
nVar <- 3
dMax <- 2
pMax <- choose(nVar+dMax,dMax)
whatTerms <- c(1,1,0,1,1,1,1,1,1,1)
 
# apply pTimEv
statio <- pTimEv(TS[,2], nVar, dMax, TS[,1], whatTerms = whatTerms, 
                 wlength = 1000, onestep = 20, removeExtr = 0.15)
# Plot the results
dev.new()
  layout(matrix(1:12, nrow=4, ncol=3, byrow = TRUE))
  what <- which(statio$whatTerms!=0)
  for (i in what) {
      plot(statio$TSdate[statio$W==1], statio$slidingoutGM[statio$W==1,i],
           xlab='TSdate', ylab='coeff', main=poLabs(nVar,dMax)[i])
     }
  plot(statio$TSdate[statio$W==1], statio$slidingoutGM[statio$W==1,pMax+1],
       xlab='date', ylab='Epsilon', main='Resid', log = 'y')