Message-ID: <Pine.LNX.4.61.0505121714410.31239@gannet.stats>
Date: 2005-05-12T16:16:54Z
From: Brian Ripley
Subject: correlogram in spatial producing values outside [-1,1]
In-Reply-To: <D71DD12EF4EDE740A09A929F6D2C509910BB77@mailuk1.rms.com>
It can happen. The covariance is produced from a set of pairs, and the
variances from all the points. It usually indicates that too fine a
binning has been used or that there is a trend in the data which has not
been removed.
On Thu, 12 May 2005, Kechi Nzerem wrote:
> I'm using the correlogram function in the spatial library to calculate
> spatial correlograms of radar data. However, I'm finding that the
> resulting values are often outside the range [-1,1], usually only at
> larger distances of separation.
>
> I'm not sure whether to be overly concerned about this, or dismiss it as
> some artefact of the data. Has anyone had similar experiences? Or know
> why this might be happening?
--
Brian D. Ripley, ripley at stats.ox.ac.uk
Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/
University of Oxford, Tel: +44 1865 272861 (self)
1 South Parks Road, +44 1865 272866 (PA)
Oxford OX1 3TG, UK Fax: +44 1865 272595