![[Back]](/images/prevpage.gif)
![[Index]](/images/index.gif)
![[Help]](/images/help.gif)
![[MSI]](/images/msi.gif)
![[ANU Online]](/images/online.gif)
Research Report SRR97-006
On the estimation of poles in intensity functions
Edwin Choi and Peter Hall
Abstract:
Motivated by spatial data on earthquakes, we
consider the problem of estimating properties of poles in point-process
intensity (and probability density) functions. Our methods are
semiparametric, requiring only "asymptotic" models for the intensity.
They produce estimates of the location and strength of poles. Strength
is expressed in terms of an exponent of regular variation, and is simply
related to the correlation dimension of the underlying point process.
It is argued that existing methods for estimating pole strength are
restrictive in terms of the range of strengths that they allow. They are
statistically consistent only in that half of the range which corresponds
to relatively strong poles, and perform poorly for poles whose strengths
lie in the middle of the range. Application of our alternative methods
to real data sets suggests that in practice, pole strengths often lie
the middle range. Our methods may also be used to estimate pole lines,
and to track the migration of pole location with time.
This service is maintained by the
Mathematical Sciences Institute (MSI)
Comments to
webmaster@maths.anu.edu.au
URL: http://wwwmaths.anu.edu.au/