dependent variable lagged

[startz]

A user has reported what looks to me to be a bug here: viewtopic.php?f=4&t=6536

I did the the following test.

Code: Select all

create u 100
series x=nrnd
ls x c x(-1)
ls x(-1) c x

Unlike the OP, I got an F-stat. However, note that I got different R^2, which is wrong unless it's round-off error.

Capture.PNG (22.75 KiB) Viewed 11239 times

Capture2.PNG (22.86 KiB) Viewed 11239 times

[EViews Gareth]

In the first you're regressing X on its future value, in the second you're regressing X on its past value.

Different estimations.


Perhaps you meant an example such as:

Code: Select all

create u 100
series x=nrnd
ls x c x(1)
ls x(-1) c x


Which also gives different R^2. That's because the sample is different.

[startz]

In the first you're regressing X on its future value, in the second you're regressing X on its past value.

Different estimations.

Gareth,

I don't think so.Both equation windows show the same sample.

And in a bivariate regression the R^2 should be the same whether you regress y on x or x on y. (The square of the correlation coefficient.)

Perhaps this is an issue of exactly how the R^2 is calculated, since the means of the dependent variables will be slightly different?

[EViews Glenn]

There is something strange but I'm not sure what it might be. If I had to bet I think it's the fact that the descriptive statistics aren't computing the lagged dependent properly. I'll look into it.