EViews.com

Itamar wrote:Ok, following Gareth's advice, I added a command option for the rtadf add in (in a quick and dirty way but it seems to work). Note that this is still experimental and it might have some bugs I missed. For it to work, you will need to replace the current rtadf.prg and rtadfsim.prg files which are in your EViews Addins\RTadf library with the attached new files.

Here is a short guide (Later on I will update the user guide):

Options:

model={1,2,3,4}; where 1=adf, 2=radf, 3=sadf, 4=gsadf
trim={}; is the window size (in fraction trems or in no. of observations)

const/trend/none; specification of the deterministic terms in the unit root test equation

info={fixed,aic,hqc,sic,maic,mhqc,msic}; information criteria for the lag length of the unit root test equation
lag={}; maximal lag length for the test equation (if "fixed" is chosen, this will be the fixed lag length)

sim; preform MC simulation for the critical values (this will prompt the simulation dialog once the calculation of the statistic is over)
graph; this will generate a graph of the results
print; this will print the spool with the result and the graph

Example:

The command,

y.rtadf(model=3, trim=0.3, trend, info=fixed, lag=1, graph, print)

will run a SADF test on y where the initial window size is set to 0.3 with a test equation specified with a constant and a trend and with one fixed lag. In addition, it will show a spool which holds the test statistic and a graph.

One caveat I have here is that, at least for now, the command cannot perform the test and the simulation automatically. However, if you wishe to preform the same test to several series having the same sample size and the same window, all you need to do is to do the simulation once, keep the critical values, and then run the same command for the remaining series (with the "sim" option shut off) via a loop.

I'll be happy to get your comments!

Enjoy bubbling...

PS, the addin now enables the user to specify the deterministic terms of the test equation, both in the test dialog and in the simulation dialog. In addition, the add in now calculates the p-value of the test statistic (according to the simulated distribution of the test statistic), but this is still an experimental feature.

I see several posts on the Eviews forum mentioned the "TStat_max_max is not defined" error, which I encountered myself too. After looking at the codes, I find the Matlab function says "function [ tstat_max, radf_cv ] = SIM_RADF( obs, win, rep, d, eta ,beta )", while in the RTADFSIM_MATLAB file, it says "xrun "[tstat_max sadf_cv]=SIM_SADF(T,t1,rep,delta,eta,beta,adfspec);"". So, there is one more input in the .prg file than in the .m file, namely the "adfspec". When I delete "adfspec" in the .prg file, everything works fine.