EqBootstrap (bootstrap standard errors)

[EViews Gareth]

This thread is about the EqBootstrap Add-in that bootstraps a linear least squares equation, allowing you to bootstrap either the standard errors or the point estimates. Note this add-in should not be used on weighted least squares or equations with ARMA terms.

[EViews Gareth]

Note for equations with auto-series (i.e. expressions such as log(x) or x^2) as variables, the bootstrap variables type of bootstrap will only work if your version of EViews 7.1 is dated at 2010/05/11 or later. If it isn't (and you can check by clicking on Help->About EViews), you should download the latest patch.

[Erik]

Good Morning,
I tried to install the add-in (I am using Eviews 7 and it was the first add-in I tried to install). After opening the file in Eviews a window opens (Binary Read - Step 1 of 3). It is not clear to me how I should proceed from there (or whether I should get to this window at all). I tried another add-in, with the same result. Can someone help?
Many thanks in advance,
Erik

[EViews Gareth]

You need EViews 7.1

[javiersan]

Small typo in "repititions"?

[EViews Gareth]

Indeed. Just fixed it and uploaded a new one.

[dvd.ceccarelli]

Dear Gareth,
I have to estimate a nonlinear LS with explicited coefficients:

d(dlp2) = c(1) + c(2)*(dlp2(-1)- (c(3)*lulc(-1)+ c(4)*mel(-1))) + c(5)*d(dlp2(-1)) + c(6)*d(lulc(-1)) + c(7)*d(mel(-1))

I have seen that bootstrao Add-in does not work. Is there a solution to run the bootstrap procedure? Thank you.

[EViews Gareth]

The add-in currently only supports linear equations.

[ii69]

Can anybody help me with one equation? It is pretty simple and I try to do boostraping for coefficients, yet I get a strange message: matrix or vector given non-positive dimension.

I have done the following:

set per 1999:1 2010:3
equation eqgexpo_ita.ls gexpo_ita grevo_ita
eqgexpo_ita.eqbootstrap(10000,SE,G,RES) gexp_ita grevo_ita

What I am doing wrong? How can I use this procedure properly?

[EViews Gareth]

You're not setting the number of bootstraps. You need to use the bs= option.

Edit: although that will fix your problem, it seems there is another issue in cases where you only have 1 regressor (such as your case). Although the bootstraps are still performed, the graph output will fail. We'll get that fixed at some point.

[EViews Gareth]

Just uploaded a new version that fixes the single regressor issue.

[yxr]

Hi, there
I just want to ask how to carry out bootstrap simulation on GARCH model. since changing "LS" to "ARCH" in this programming does not work.

Thanks in advance

[EViews Gareth]

As you noted, the Add-in doesn't support GARCH, so you'll have to program one yourself.

If all you want to do is bootstrap the exogenous variables in the GARCH estimation, it should be relatively straight forward. Anything more complicated will be, well, more complicated

[yxr]

All right.
Thanks