[R] about contour - get contour coordinates - exclude area display

[Prof Brian Ripley]

On Sat, 31 Jan 2004, Patrick Giraudoux wrote:

> Dear all,
> 
> I wonder about what could actually be possible with the function "contour":
> 
> 1/ - the definition of contour lines is most often meaningless when the
> contours are drawn in areas where no real data points exist. It can
> however happen that irregular distributions lead to more or less
> irregular clouds of data points. Interpolations (eg: loess regression,
> GLM, etc...) are however generally made on regular grids passed to the
> function contour and thus can cover areas with no/few data points to
> some extent. It would be most useful to display the contours with a
> polygon argument allowing the exclusion of those areas (this is
> possible, for instance, with controur.krige of the package GeoR).  Is

There is no function cont(r)our.krige in geoR. There is a function
`contour.kriging', and that just calls contour, so you too can do whatever
you think that does (and the help page does not tally with your
description).

> there a way to exclude the display of contour lines out of polygon
> coordinates passed to the contour function (or another function in a
> package doing this) ?

Contour works on a rectangular grid.  Set to NA those cells where you have
no data, and you will not get contours there.  Simple!

> 2/ - it would also be most useful to get the contour lines coordinates
> (for instance, for importation in a GIS). Is there a way to get them?


Paul Murrell has fulfilled that request a while back.  See the clines
package on CRAN, or contourLines in the development version of R.

-- 
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