{smcl}
{* *! version 1.0 2008-02-18}{...}
{cmd:help newsimpact}
{hline}
{title:News impact curve for ARCH models}
{title:Syntax}
{p 8 25 2}
{cmd:newsimpact}
[{cmd:,}
{opt s:igma2(a)}
{opt r:ange(b)}]
{title:Description}
{p}{cmd:newsimpact} is for use after {cmd:arch}.
It plots the news impact curve, the response in the conditional variance,
sigma^2(t),
to an innovation in the standardized error term, z(t-1).
When calculating the response in sigma^2(t) historical conditional variances
(sigma^2(t-i), i > 0) are set to sigma^2 and historical error terms (z(t-i), i > 1)
are set to 1. The default value for sigma^2 is an estimate of the unconditional variance,
the mean of the estimated conditional variances. Finally error terms that enter in
the conditional variance formula are obtained as e(t) = sigma*z(t).
{title:Options}
{synoptset 15}{...}
{synopthdr:option}
{synoptline}
{synopt:{opt s:igma2(a)}}Use sigma^2 = a instead of the default. The
default is to set sigma^2 to the mean of the estimated conditional variances.
{synopt:{opt r:ange(b)}}Plot the news impact for z(t-1) = -b to +b instead of
the default z(t-1) = -2 to +2.
{synoptline}
{title:Examples}
{pstd}Setup{p_end}
{phang2}{cmd:. webuse wpi1}{p_end}
{pstd}Fit GARCH(1,1) and plot news impact curve{p_end}
{phang2}{cmd:. arch D.ln_wpi, arch(1) garch(1)}{p_end}
{phang2}{cmd:. newsimpact}{p_end}
{pstd}Fit EGARCH(1,1) and plot news impact curve (yes, the curve is weird){p_end}
{phang2}{cmd:. arch D.ln_wpi, earch(1) egarch(1)}{p_end}
{phang2}{cmd:. newsimpact}{p_end}
{title:Author}
{pstd}Sune Karlsson, Örebro University, Sweden{p_end}
{pstd}sune.karlsson@oru.se{p_end}