🚀 go-pugleaf

In article <2A817F03.29491@noiro.acs.uci.edu> manderse@orion.oac.uci.edu (Mark Andersen) writes:
>I have been asked to help analyze some data being collected on desert
>tortoises. Among other things, a goal of the study is to estimate the 
>number of tortoises living in a 70 sq. mi. area in the Mohave desert
>here in southern California. Data are being collected on the number
>of "tortoise signs" (i.e., tracks and feces) in fixed triangular 
>transects (four per square mile); these data will be converted to
>numbers of tortoises using standard conversions.
>My idea was to fit an interpolated surface of tortoise population
>density (via kriging), then integrate this surface over the study
>area to obtain an estimate of tortoise numbers. Obviously, there's
>more to this than just what I've presented here, but this is the 
>crux of my question. Is it considered OK to use the integral of an
>interpolated surface as an estimate of population size in a study region
>in this way (seems reasonable to me)? 
>
>Mark Andersen                   manderse@orion.oac.uci.edu
>Dept. of Ecology and Evolutionary Biology

This is a natural application for block kriging rather than the
point kriging you seem to have in mind.  The attraction of
using block kriging directly rather than the approach you 
outline is that you will be able to get a standard error
to go with your estimate of the total numbers.  You will not
get this by using point kriging and integrating the result.

Doug Hawkins
------------
    ##########        Douglas M Hawkins          Dept of Applied Statistics
    ##  ##  ##                                      University of Minnesota
######  ##  ######    Voice  (612) 625-8159        352 COB, 1994 Buford Ave
##  ##  ##  ##  ##    Fax    (612) 624-2719                         St Paul
######  ##  ##  ##    Internet:  doug@umnstat.stat.umn.edu    MN 55108-6042