Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: Out-of sample ARMA forecast error variance for confidence bands


From   Michael Bechtel <michael.bechtel@ir.gess.ethz.ch>
To   <statalist@hsphsun2.harvard.edu>
Subject   st: Out-of sample ARMA forecast error variance for confidence bands
Date   Wed, 29 Jul 2009 14:41:54 +0200

Dear Statalist members

is there a convenient way to obtain confidence bands for ARMA (out-of sample) forecasts in Stata? Computing this for a one-step ahead forecast is simple, but more steps get complicated and also strongly increase in complexity when using ARMA models with several AR and MA terms. Any suggestions are greatly appreciated!

Many thanks in advance!

Michael Bechtel

PS: I read Robert Yaffee's reply back in 2006, but unfortunately this advice only applies to one-step ahead forecasts (http://www.stata.com/statalist/archive/2006-06/msg00335.html )

--
ETH Zürich
Dr. Michael M. Bechtel
Swiss Federal Institute of Technology Zurich
Center for Comparative and International Studies
WEC C 25
Weinbergstrasse 11
8092 Zürich

michael.bechtel@ir.gess.ethz.ch
http://www.ib.ethz.ch/people/mbechtel

+41 44 632 62 68 Phone
+41 44 632 12 89 Fax
--------------------------------------------------------------

Am 29.07.2009 um 14:14 schrieb Giorgia Maffini:

Dear Statalist Members,

I wonder if anybody could kindly help me out with the following problem.

I am using a difference-GMM estimator (Arellano and Bond (1991), Review of Economic Studies - AB) employing -xtabond2-. Instead of first- differencing, I would like to second-difference the equation and instrument some of its
endogenous covariates with suitable lags of their own levels.

In other words, the dependent variable in the newly differenced equation would need to be D2.y = y_(it)-y_(it-2) [instead of D.y = y_(it)-y_(it-1), as in the standard AB framework]. The endogenous covariate would be D2.x = x_(it)-x_(it-2) [instead of D.x = x_(it)-x_(it-1), as in the standard AB framework]. I would instrument D2.x with the levels x_(it-3), x_(it-4), etc. and other suitable
instruments (e.g. z_it).

The following command does NOT seem to produce the requested procedure:

xi: xtabond2 D2.y L.D2.y D2.x , /*
*/ gmm( y  x , lag(3 4) collapse ) iv( z , passthru ) /*
noleveleq two robust

Does anybody know how to implement the aforementioned estimation using -xtabond2- ?

The version of Stata that I am using is the following:

Stata/MP 10.1 for Windows 64-bit x86-64
Born 02 Feb 2009

Thank you for your consideration.

Regards,

Giorgia

-
Giorgia Maffini - Research Fellow
Oxford University Centre for Business Taxation - Said Business School
Park End Street, Oxford OX1 1HP
Tel: 01865 614847

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index