In random sampling, any member of the population has an equal chance of being selected to contribute to the sample. In practice, this means that the set of potential sample units are identified and then the individuals that are actually sampled are selected using a randomization technique, such as throwing a dice, flipping a coin, or using a random number table. For example, 100 contiguous 1m2quadrats could be identified along a 100m tape, and then 20 of these quadrats are selected from a random number table and measured. Similarly, 10 potential transects could be systematically identified at 10m intervals along a 100m baseline, and 3 of these transects selected for sampling using a deck of cards.
Random selection of sample units is an underlying assumption of most statistical inference techniques, because it ensures that the sample unit selection is free from personal bias and not confounded by possible spatial patterns within the vegetation. Therefore, a major advantage of adopting random sampling is that data sets can be compared using conventional statistical inference techniques that estimate the sample mean and its precision.
However, random sampling is often impractical to apply in its purest sense. For example, randomly located quadrats are often difficult to accurately locate, particularly when an extensive area is sampled, and a large proportion of sampling time is devoted to locating the quadrats. In addition, a larger sample size is needed to obtain adequate precision under random sampling, because the technique ensures that all the variability of the population is represented in the sample. Therefore, simple random sampling designs usually feature low sampling efficiency. Finally, random selection processes may lead to an uneven distribution of sample units across the site, so that some areas may be poorly represented during sampling.
Several designs have been developed to address these disadvantages while maintaining the principles of randomization, including restricted random techniques (arrangement of sample units within blocks or clusters) and stratified sampling.
References and Further Reading
(Note: pdf files require Adobe Acrobat (free) to view)
Cook, C.W., and J. Stubbendieck. (eds). 1986. Range research: Basic problems and techniques. Society for Range Management, Denver, CO. pp 221-223.
Daubenmire, R. 1968. Plant communities: A textbook on plant synecology. Harper Row, New York, NY. pp 81-86.
Dowdy, S. and S. Weardon. 1991. Statistics for research. John Wiley Sons, New York, NY. 2nd ed. pp 23-25.
Osborne, J.G. 1942. Sampling errors of systematic and random surveys of cover-type areas. Journal of American Statistics Association 37:257-264.
Salmon, S.C. 1953. Random versus systematic arrangement of field plots. Agronomy Journal 45:459-462.
Wester, D.B. 1992. Viewpoint: replication, randomization, and statistics in range research . Journal of Range Management 45:285-290. (pdf)