tsDyn (version 11.0.4.1)

TVECM.SeoTest: No cointegration vs threshold cointegration test

Description

Test the null of no cointegration against threshold cointegration with bootstrap distribution of Seo (2006)

Usage

TVECM.SeoTest(
  data,
  lag,
  beta,
  trim = 0.1,
  nboot,
  plot = FALSE,
  hpc = c("none", "foreach"),
  check = FALSE
)

Value

A list containing diverse informations:

Estimated threshold parameters and usual slope parameters.

Value of the test.

Critical and Pvalue from bootstrap distribution.

Arguments

data

time series

lag

Number of lags to include in each regime

beta

Pre-specified cointegrating value (i.e. cointegrating vector will be c(1, -beta))

trim

trimming parameter indicating the minimal percentage of observations in each regime

nboot

Number of bootstrap replications

plot

Whether a grid with the SSR of each threshold should be printed

hpc

Possibility to run the bootstrap on parallel core. See details in TVECM.HStest

check

Possibility to check the function by no sampling: the test value should be the same as in the original data

Author

Matthieu Stigler

Details

For this test, the cointegrating value has to be specified by the user.

The model used is one where the threshold effect concerns only the cointegrating vector, and only in the outer regimes.

Due to the presence of parameters unidentified under the null hypothesis, the test employed is a Sup-Wald test, that means that for each combination of the thresholds, a Wald Test is computed and the supremum of all tests is taken. For each bootstrap replication, this approach is taken, so that the test is really slow.

References

Seo, Myunghwan, 2006. "Bootstrap testing for the null of no cointegration in a threshold vector error correction model," Journal of Econometrics, vol. 127(1), pages 129-150, September.

See Also

TVECM for estimating a TVECM, TVECM.sim for simulating/bootstrap a TVECM,

Examples

# As the function takes long long time to be executed, we show in in don't run environement
if (FALSE) {
data(zeroyld)

#can be useful to check whether the bootstrap is working: 
#without sampling, results of boot should be same as original
#this is indeed not always the case duye to floating point algorithm
TVECM.SeoTest(zeroyld,lag=2, beta=1, trim=0.1,nboot=2, plot=FALSE,check=TRUE)

#then run the function:
TVECM.SeoTest(zeroyld,lag=2, beta=1, trim=0.1,nboot=100, plot=FALSE,check=FALSE)
}