Exporting CUSUM Data

[brennan6738]

Does anyone know if there is a way to export the actual data from a CUSUM stability test? I need to chart it in Excel and I can't just export the chart from Eviews.

[trubador]

I am not sure if it can be done via a single command. However, the following piece of code should be doing what you want:

Code: Select all

call cusum("eq01") 'Please enter the name of your equation in double quotes. subroutine cusum(string %eq) !nobs = {%eq}.@regobs !ncoef = {%eq}.@ncoef !df=!nobs-!ncoef {%eq}.rls(q,s) series cusum=@cumsum(r_res/{%eq}.@se) smpl @first + !ncoef @last series ubound = .948*@sqrt(!df)*(1+2*(@trend-!ncoef+1)/!df) series lbound = -ubound smpl @all endsub

[brennan6738]

Data and program.

[trubador]

You must not change any part of the subroutine. And there has to be a line in the program window, which calls the subroutine. The correct way is as follows:

Code: Select all

call cusum("multivariate") subroutine cusum(string %eq) ... endsub

Please do not hesitate to refer to programming manual or users guide for additional information...

[tcfoon]

Dear Trubador,

I am interested with the CUSUM programme you posted to the forum. Could you please share with me the programme for CUSUM of Squares (CUSUMSQ) because I intent to change the significant level from 5 to 10 per cent. You are using 0.948 as 5% why not 0.95, while if I wish to use 10 per cent does it mean that I should you 0.90 or 0.898?

Please advice..

Thank you,

[trubador]

The parameter values of CUSUM test should be 1.143, 0.948 and .850 respectively for 1%, 5% and 10% significance levels. For details please see: Brown, R. L., Durbin, J. and Evans, J. M. (1975). "Techniques for Testing the Constancy of Regression Relationships over Time", Journal of the Royal Statistical Society, vol. 37, pp. 149-192.

Calculation of confidence intervals for CUSUM of squares are different than that of CUSUM, since the parameter values also depend on degrees of freedom. At this point, you should refer to original study: Durbin, J. (1969). 'Tests for Serial Correlation in Regression Analysis Based on the Periodogram of Least-Squares Residuals", Biometrica, vol. 56, pp. 1-15.

The CUSUMSQ series can be generated as follows:

Code: Select all

series cusumsq=@cumsum(r_res^2/@sumsq(r_res))

[rgarg]

I have a question similar to the above. How can the default 5% confidence interval bands for CUSUM Squares be changed to 1% or 10%. What would be the code for it ?

rgarg