Page 1 of 1

Linear Restriction Analysis

Posted: Tue Feb 24, 2015 2:11 pm
by diggetybo
Hello,

I am interested in testing a coefficient restriction for a coefficient to be greater than a certain value, say .01.

When I type this into wald coefficient restriction in eviews 6, it says, "invalid specification"

I tried the following syntax: c(2)>.01

So the only acceptable restrictions are in "=" format?

Are inequality restrictions possible in eviews 6? If not, are they possible in eviews 8 or 9?

please advise, :)

Re: Linear Restriction Analysis

Posted: Tue Feb 24, 2015 2:29 pm
by startz
Just use the t-statistic from the test on equality, and then look up a one-tailed critical value from a table.

Re: Linear Restriction Analysis

Posted: Tue Feb 24, 2015 3:04 pm
by diggetybo
Thanks startz,

If I could follow up for a second,

I believe an F statistic would work as well right? In eviews 6, that's all the wald coefficient restriction output gives.

let's use an example, for clarity:

Say I wanted to examine the following restriction:

c(2)>c(3)

In eviews I type: c(2)=c(3)

Supposing this null was rejected, from comparing the F stat with a critical value, then we know that c(2) is either greater than or less than c(3). To test if it was greater we use a right tailed f distribution. Is this what you meant?

Re: Linear Restriction Analysis

Posted: Tue Feb 24, 2015 3:13 pm
by startz
I don't believe there is a simple way to do this with an F-statistic. However, so long a you are testing a single restriction the t-stat is just the square root of the F-stat.

Re: Linear Restriction Analysis

Posted: Tue Feb 24, 2015 3:29 pm
by diggetybo
Oh true enough,

I can work with that. Thanks so much!