optimal.time.scale: Best time-scale according to the approach by Watanabe et al....
In deanfantazzini/bubble: Tests for financial bubbles and explosive behavior

Description Usage Arguments Details Examples

This function looks for the best time-scale according to the approach by Watanabe et al. (2007a,b)

optimal.time.scale(
  xdat,
  maxlag = 10,
  min.win = 10,
  max.win = 20,
  n.replications = 2000,
  clust_number = 8,
  original = TRUE
)

`xdat`	is a T x 1 numeric data vector
`maxlag`	is the maximum lag to be used to select the optimal AR lag order with AIC
`min.win`	is the minimum window size to be tried during the simulations
`max.win`	is the maximum window size to be tried during the simulations
`n.replications`	is the number of simulated AR(p) processes (for each window size)
`clust_number`	is the number of clusters for parallel computation
`original`	if TRUE the original method by Watanabe et al. (2007a,b) is used, otherwise a simplified approach which does not required nonlinear optimization.

This function looks for the best time-scale according to the approach by Watanabe et al. (2007a,b), using either the original method or a simplified approach which does not required nonlinear optimization. The optimal windown size according to Watanabe et al. (2007a,b) is the minimum window size for which always omega(i, Ti)<=1 holds. If an estimated parameter omega(i,ti)>1, the following message is printed to the screen, "series is divergent, at least for one iteration = [", i, "], omega(i,Ti) > 1" and the function stops checking further.

 ## Not run: 
 x= arima.sim(list(order = c(3,0,0), ar = c(1.2, -0.2, -0.1)), n = 500)
 optimal.time.scale(xdat=x,maxlag=10,min.win=5,max.win=20,n.replications=2000,
 clust_number=8,original=FALSE)
 
## End(Not run)