Wildlife Software



RANDSEL: Randomly selecting and assigning 

animals to treatment groups



Robert T. Sugihara



Author's address: U.S. Department of Agriculture, Animal 

and Plant Health Inspection Service, Animal Damage Control, 

National Wildlife Research Center, P.O. Box 10880, Hilo, HI 

96721-5880, USA.



Key words: animal selection, random assignment, 

randomization, treatment groups



	Efficient, random selection and allocation of animals 

to experimental treatment groups is a prime concern for 

experimental biologists and statisticians. The Animal 

Welfare Act and U.S. Environmental Protection Agency Good 

Laboratorv Practice regulations have elevated aspects of 

these statistical concerns to regulatory issues (Engeman and 

Shumake 1993). Computerized assignment of animals to 

treatment groups greatly facilitates meeting these 

statistical and regulatory requirements (Sugihara 1992).  

	RANDSEL is an interactive computer program developed at 

the National Wildlife Research Center's  (NWRC) Hawaii Field 

Station as a quick and easy method for assigning animals to 

treatment groups while addressing statistical concerns and 

meeting regulatory record-keeping requirements. RANDSEL is 

used to search a dataset of quarantined animals to select 

animals for testing based on user-specified criteria for 

species, sex, and weight ranges. To avoid confounding 

treatment effects with those due to animal weights,  the 

program blocks animals by weight groups prior to assigning 

them to treatment groups. This restriction on randomization 

should be considered when selecting appropriate analysis 

methods. The random-number generator (Bratley et al. 

1983:180-213), which produces a sequence of pseudo-random 

numbers in the selection and randomization process, has been 

extensively tested for accuracy and nonperiodicity (Brody 

and Morais 1987). 

	Users can import data from an existing database file 

(ASCII text format) or enter it interactively when executing 

the program. Other variables besides species, sex, and 

weight may be specified to control record selection and 

treatment assignment. The program will accept data 

consisting of 2 categorical variables, a record-

identification field, and a measured continuous variable. 

Records can be entered in any because RANDSEL automatically 

sorts the records. If an existing database file is used, a 

header describing the data format must be placed at the top 

of the input file. Each record must appear on a separate 

line. The current program can accept <=5,000 records, 

however, this may be increased based on available computer 

(RAM) resources.

	After data is entered, the user specifies the desired 

selection criteria (e.g., species, sex, weight limits),  the 

number of treatments (NT), and the number of animals 

(replicates)/treatment (NR). For each species and sex 

combination, the user can select a lower and upper body-

weight limit (e.g., male Norway rats >100 g and < 200 g). 

The user then selects a random number seed to initiate the 

pseudo-random number generator and start the random 

selection. 

	RANDSEL randomly selects the appropriate number of 

animals, based on the selection criteria, NT and NR, from 

the pool of available animals. It then ranks these animals 

by weight, from the lightest to the heaviest, and divides 

them into NR weight groups.  The program then randomly 

assigns a treatment to each individual in each weight group. 

If the required number of animals (NT x NR individuals) are 

not found in the pool, the program prompts the user to 

repeat the process with a different set of selection 

criteria or exit the program. 

	For each run, RANDSEL reports: (1) selection criteria, 

(2) a list of animals that meet the selection criteria, (3) 

the number of and a list of the randomly selected animals, 

and (4) a list of randomly selected animals with assigned 

treatments. The user can then choose another sex or species-

sex combination and execute the selection process again or 

exit the program. Optionally, the user can print program 

results.

	Researchers at the NWRC Hawaii Field Station have found 

that RANDSEL significantly reduces the time required and 

number of errors involved in selecting and assigning animals 

to experimental treatment groups. They can, for example, 

quickly select 10-20 males and females from each of 3 rat 

species this for feeding bioassays involving 3-4 different 

baits, randomly assign them to treatment groups, and obtain 

complete printed documentation in <1 minute.

	RANDSEL can be installed on any standard IBM-compatible 

microcomputer running the MS-DOS or OS/2 operating system. 

The program is written in FORTRAN 77, and the compiled 

executable file uses 228K bytes of RAM. A math coprocessor 

is recommended for accessing large datasets. Data files 

created with other database software and converted to 

standard ASCII format can also be used. The executable 

program file, an example input animal data file, and a 

user's manual will be provided upon request by sending a 

formatted 3.5 - or 5.25 - inch floppy disk and a stamped 

self-addressed mailing container to Robert T. Sugihara, 

USDA/APHIS, National Wildlife Research Center, Hawaii Field 

Station, P.O. Box 10880, Hilo, HI 96721. RANDSEL is also 

available through the Bird Monitor bulletin board (301-497-

5831), FTP site: ftp.im.nbs.gov in /pub/software/CSE, or 

World Wide Web site at http://www.im.nbs.gov/tws/cse.html.



	Acknowledgments. I thank M. E. Tobin, project leader, 

for providing the opportunity and resources  to develop this 

and other computer applications that have greatly simplified 

and improved the efficiency of research operations at the 

NWRC Field Station. M. E. Tobin, R. M. Engeman, and A. E. 

Koehler suggested improvements and critically reviewed 

earlier drafts of manuscript.



Literature cited

BRATLEY, P., B. L. Fox, and L. E Schrage. 1983. A guide to 

simulation. Springer-Verlag, New York, N.Y. 383pp.

BRODY, M. and D. MORAIS. 1987. UNIF is a fast, reliable 

random number generator for IBM-AT microcomputer U.S Fish 

and Wildl. Res. Info. Bull. No. 87-84.

ENGEMAN, R. M., AND S. SHUMAKE. 1993. Animal welfare and the 

statistical consultant. Am. Stat. 47(3):229-233.

SUGIHARA, R. T. 1992. Use of microcomputers for randomly 

assigning animals to treatment groups, formulating baits, 

and keeping records. Proc. 15th Vertebr. Pest Conf. 15:249-

251.



Robert (Bob) T. Sugihara is a Wildlife Technician with the 

U.S. Department of Agriculture's National Wildlife Research 

Center's Hawaii Field Station based in Hilo, Hawaii. He 

received a B.S. in Tropical Agriculture from the University 

of Hawaii at Manoa. Bob has 20 years of wildlife research 

experience with the U.S. Departments of Interior and 

Agriculture, focusing on rodent damage problems in Hawaiian 

agricultural crops and conservation areas. He has a special 

interest In microcomputers and the development of computer 

applications to increase research productivity.



Software Editor: Smith