Program CENTROID User Instructions
Program CENTROID is used to test the hypothesis that two samples belong to the same bivariate
distributions. This procedure is described in Mardia (1967:J. Roy. Stat. Soc., Ser. B,
29:320-342), and Batschelet (1972: Pages 61-91, in Animal Orientation and Navigation, NASA
SP-262).
CENTROID will read either recovery records or X-Y coordinates. Recovery records, which
include recovery latitude and longitude, of the two samples are read into the program. The
program will print out the average latitude and longitude of each group, the average latitude and
longitude of the combined sample, the number of duplicate recovery locations, Mardias test
statistic, and the probability level.
In order to account for duplicate recovery locations the program prints out four test statistics.
The first statistic is computed by breaking all "ties" in favor of group 1 (i.e. group 1 recovery
location is ranked lower than group 2). The second statistic is computed by breaking all ties in
favor of group 2. The third and forth statistics are computed by averaging the ranks of all
duplicate recovery locations.
Two input files are required by this program:
1) Recovery file for group 1.
2) Recovery file for group 2.
These recovery files contain the lat-long of recovery in columns 33-39 in the following format:
Cols 33-34 : Latitude degrees. 35 : Latitude ten-minute block. 36-38 : Longitude degrees. 39 : Longitude ten-minute block.
If the data are in X-Y coordinate form, the coordinates are read in 'free' format (ie: x,y).
CENTROID will prompt the user for the names of the two input files and the output file and
titles.
Since the number of duplicate locations has an extreme effect on the test statistic, the program
displays the number of ties when recovery locations were ranked. CENTROID also displays the
highest order tie (MAX TIES COMPUTED), followed by a breakdown of all other ties (1 WAY
TIES, 2 WAY TIES, ...). A "1 way tie" is a distinct location. A "2 way tie" is when two and only
two observations occur in the same location. A "mixed tie" occurs when observations from both
groups are involved in a tie.
The test statistic is computed four different ways and printed out with a short explanation of how
ties were broken in each case.
Program RUNCENT.BAS runs program centroid for all possible combinations of data files (useful for comparing distributions from different reference areas).