Thanks, scott. I tried the command several times later, the error of
not predicting did not happen again. That has me wondering why I see
it in the first place. Now every time I try, the prediction would be
for both in and out of sample. As I could not reproduce the mistake, I
could not figure out what went wrong that time. Next time, I would try
more times before surrending and ask for help!
Thanks, all!
Faye
On 5/5/05, Scott Merryman <[email protected]> wrote:
> Do you have observations for variable x beyond 1990? If so then -predict-
> should give you out of sample predictions. In general, -predict- calculates
> the requested statistic for all possible observations, whether they were
> used in the model or not.
>
> In the example below, the model is estimated on data up to 1980q1, but the
> predicted values go to 1990q4.
>
> . use http://www.stata-press.com/data/r9/wpi1.dta, clear
>
> . arch ln_wpi if tin(,1980q1), arch(1) garch(1) nolog
> BFGS stepping has contracted, resetting BFGS Hessian (0)
>
> ARCH family regression
>
> Sample: 1960q1 to 1980q1 Number of obs = 81
> Wald chi2(.) = .
> Log likelihood = 85.18144 Prob > chi2 = .
>
> ----------------------------------------------------------------------------
> | OPG
> ln_wpi | Coef. Std. Err. z P>|z| [95% Conf.
> Interval]
> -------------+--------------------------------------------------------------
> ln_wpi |
> _cons | 3.423398 .0011066 3093.53 0.000 3.421229
> 3.425567
> -------------+--------------------------------------------------------------
> ARCH |
> arch |
> L1. | 1.137414 .630359 1.80 0.071 -.0980673
> 2.372895
> garch |
> L1. | .0000419 .1747459 0.00 1.000 -.3424538
> .3425376
> _cons | 7.66e-06 3.41e-06 2.25 0.025 9.76e-07
> .0000144
> ----------------------------------------------------------------------------
>
> . predict y, y
>
> . l in -5/l
>
> +--------------------------------------+
> | wpi t ln_wpi y |
> |--------------------------------------|
> 120. | 109.4 1989q4 4.695011 3.423398 |
> 121. | 111 1990q1 4.70953 3.423398 |
> 122. | 110.8 1990q2 4.707727 3.423398 |
> 123. | 112.8 1990q3 4.725616 3.423398 |
> 124. | 116.2 1990q4 4.755313 3.423398 |
> +--------------------------------------+
>
> Scott
>
> > -----Original Message-----
> > From: [email protected] [mailto:owner-
> > [email protected]] On Behalf Of F. Gao
> > Sent: Wednesday, May 04, 2005 10:20 PM
> > To: [email protected]
> > Subject: st: How to let Arch give out of sample prediction??
> >
> > Dear all,
> >
> > I am using arch command to estimate a time series for a subset of
> > years, and hoping to get predicted values for fitted values as well as
> > the conditional variance out of sample (after 1990). I only get the
> > in-sample predictions somehow. Here is what I did
> >
> > arch y x if year<1990, archm arch(1) garch(1)
> >
> > predict yhat
> >
> > predict cond_var, variance
> >
> > Can anyone tell me why the last two commands would not give out of
> > sample predicitons? Thanks!
> >
> > Faye Gao
> >
>
> *
> * For searches and help try:
> * http://www.stata.com/support/faqs/res/findit.html
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
>
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
