Slope Dummy Variables Error Message NEAR SINGULAR MATRIX

[thomas84]

Dear eviews community,

I am having an issue using a slope dummy variable in my time series regression. I create this variable by multiplying my variable of interest (total government expenditure, 1960-2013) with a time dummy to test for a structural break in 1980. Hence, my time dummy has the value 0 before 1980 and the value 1 after 1980.

Now, having included this new variable into the regression I get the following error message: "NEAR SINGULAR MATRIX ERROR. REGRESSORS MAY BE PERFECTLY COLLINEAR".

Could you please let me know how to overcome this problem, and what to do with the regression. i.e. whether to include only the slope dummy in the regression?

Thank you.

Best
Thomas

[startz]

You might want to show us exactly the commands you used and perhaps post your EViews workfile.

[thomas84]

Thank you for your response. I did not use any specific command (e.g. @expand to create categorical dummy variable). I simply added an additional variable to the model which is the multiplication of my variable G with a time dummy D80(0 before 1980, and 1 after 1980). The command line is the following:

dlogi c dlogg dlogrs dlogg*d80

So I simply put in the new variable into the equation and get the error message regarding perfect collinearity. Does this clarify? Should I leave out my variable dlogg then? However, I wanted to test for both.

Thank you.

[startz]

That command should work. You might want to post your workfile, including the equation you are estimating.

[thomas84]

Sure this is the equation:

dlog(i15) c dlog(yp15) dlog(gcum15) dlog(gcum15*d80) log(i15(-1)) log(yp15(-1)) log(gcum15(-1))

I attach workfile.

[startz]

The problem is that by multiplying by the dummy variable you are taking logs of zero for many observations. Maybe you meant

Code: Select all

dlog(gcum15)*d80

rather than

Code: Select all

dlog(gcum15*d80)

[thomas84]

Yes thank you, silly mistake. Thanks a lot.