Bill Shipley. 1993. A null model for competitive hierarchies in competition matrices. Ecology 74:1693-1699.
Supplements
Supplement 1: Fortran code for assessing transitivity
Ecological Archives E074-002-S1 (ESPS 9302)Files available for download
complet.exe
complet.for
rcomplet.exe
rcomplet.for
readme.txtDescription
Instructions for use of COMPLET and RCOMPLET
These FORTRAN programs accompany the paper:
Shipley, B. 1993. A null model for complete transitivity in competition matrices. Ecology 74:1693-1699.
The program COMPLET will search through binary competition matrices, identify all paths within the matrix that are (completely) transitive, and calculate the probabilities of finding at least that many transitive paths under the null hypothesis of independence of interactions. The probabilities are estimated by a Monte Carlo simulation. In the program, you are asked how many species are in the matrix,and then asked to input the binary matrix one row at a time with a single space between entries. Thus, the first row of a matrix with 5 species might be:
0 1 1 0 1
Once the matrix has been entered, the program asks you to give a file name to store the results (DOS conventions), identifies all transitive paths, and prints them to your output file. Next, the program asks you how many iterations you want to use in the Monte Carlo simulation. A small number (1000) will take only a short time to calculate, but the estimated probabilities will be less precise. The actual time required for the Monte Carlo estimates will depend on the number of species in the matrix, the number of iterations, and the speed of your computer. On a 486, 33 Mhz machine, 10,000 iterations of a 4 species matrix takes less than a minute. 10,000 iterations of a 13 species matrix may take almost an hour.
For most practical uses, 10,000 iterations will provide more than enough accuracy and 1000 iterations are probably enough. Also included is a program that performs Monte Carlo simulations based on the null model for a given number of species (RCOMPLET), and outputs the full probability distribution. You DON'T need to run this program if you only want to calculate significance levels, since the main program - COMPLET - does this automatically.
Finally, these programs are offered free of charge. The FORTRAN source code is also given (the *.FOR files). If you modify these programs, please insert your name, address and telephone number, so that subsequent users know that I am not responsible for the changes, and notify me of your changes.
Author
Dr. Bill Shipley
Departement de Biologie
University de Sherbrooke
Sherbrooke (Quebec)
CANADA J1K 2R1
(819) 821-7079
BSHIPLEY@UDESMA.BITNET
The original paper or microfiche copies of supplements submitted prior to 1998 are on file in the Publications Office of the Ecological Society of America.
ESA Publications | Ecological Archives | Permissions | Citation | Contacts