Completely Randomized Design

Application

CRD is mostly useful in laboratory and green house experiments in agricultural, biological, animal, environmental, and food sciences, where experimental material is reasonably homogeneous. It is more difficult when the experimental units are people.

Advantages and Disadvantages

This design has several advantages. It is very flexible as any number of treatments may be used, with equal or unequal replications. The design has a comparatively simple statistical analysis and [Page 200]retains this simplicity even if some observations are missing or lost accidentally. The design provides maximum degrees of freedom for the estimation of error variance, which increases the precision of an experiment.

However, the design is not suitable if a large number of treatments are used and the experimental material is not reasonably homogeneous. Therefore, it is seldom used in agricultural field experiments in which soil heterogeneity may be present because of soil fertility gradient or in animal sciences when the animals (experimental units) vary in such things as age, breed, or initial body weight, or with people.

Layout of the Design

The plan of allocation of the treatments to the experimental material is called the layout of the design.

The treatments are allocated to the experimental units or plots in a completely random manner. Each treatment has equal probability of allocation to an experimental unit.

Given below is layout plan of CRD with four treatments, denoted by integers, each replicated 3 times and allocated to 12 experimental units.

Randomization

Some common methods of random allocation of treatments to the experimental units are illustrated in the following:

Consider an experiment with less than or up to 10 treatments. In this case, a 1-digit random number table can be used. The treatments are allotted a number each. The researcher picks up random numbers with replacement (i.e., a random number may get repeated) from the random number table until the number of replications of that treatment is exhausted.

For experiments with more than 10 treatments, a 2-digit random number table or a combination of two rows or columns of 1-digit random numbers can be used. Here each 2-digit random number is divided by the number of treatments, and the residual number is selected. When the residual is 00, the divisor number is selected. The digit 00 already occurring in the table is discarded. The digit 00 is discarded.

...