The RATS Software Forum

Posted: **Fri Jul 10, 2015 12:06 am**

Hi
Does anyone know how to test the unit root in bounded series? Granger (2010) gave the idea. Cavaliere and Xu, 2014,proposed a modified ADF test for the bounded series. The title of the paper is "Testing for unit roots in bounded time series" Journal of Econometrics. http://www.sciencedirect.com/science/ar ... 7613001905 Algorithm are given in page 262.

I am wondering if there is any RATS procedure available to carry out the test.
Thanks

Posted: **Mon Jul 13, 2015 12:07 pm**

There really isn't much to that. You use a standard unit root test and you just have to simulate a critical value that's specific to the (standardized) bounds for the particular data set. This shows simulation of a bounded random walk.

Code: Select all

*
* Simulation of ADF statistic from Cavaliere and Xu(2014), "Testing for
* unit roots in bounded time series,"  Journal of Econometrics, vol.
* 178(P2), 259-272.
*
* Upper and lower bounds for standardized process
*
compute cupper=1.5
compute clower=-.5
*
* Size of Monte Carlo sample. (Needs to be >= number of data
* points---authors suggest n=T).
*
compute n=500
compute ndraws=10000
*
set adfstat 1 ndraws = 0.0
do draw=1,ndraws
   *
   * Simulate random walk bounded below by clower and above by cupper
   *
   set(first=0.0) xstar 1 n = xtest=xstar{1}+%ran(1.0/sqrt(n)),%min(%max(xtest,clower),cupper)
   *
   sstats(mean) 1 n xstar^2>>mu2 xstar>>mu1
   *
   * This is the T(rho-1) statistic.
   *
   compute adfstat(draw)=(xstar(n)^2-1)/(2*(mu2-mu1^2))
end do draw
*
stats(fract) adfstat

The RATS Software Forum

unit root test for bounded series

unit root test for bounded series

Re: unit root test for bounded series