SAMPLE SIZE DETERMINATION FOR AGRONOMIC EXPERIMENTS

Abstract

Determining sample size to estimate the mean value of a target (infinite) population is of prime importance to have correct results and conclusions. In general, sample sizes differ as the studied objects differ. Another factor affect sample size is number of replicates or groups. Small objects, such as seeds, grains, and bacterial cell sizes require in general larger sample sizes as compared to large objects, such as human patients, large animals, and large trees. In this short article, a mixture of 12 crosses, maize kernels (k) were mixed to have 12,000 kernels. Sample sizes of 50,100, 150, 200, 250, and 300k were taken to find out which sample size fits the mean of target population (12,000 k). Sample size of at least 100 k and up gave the best good fit as compared by Chi-square, P=0.10 and 0.05 at one of degree of freedom. This result was on a heterogenous population of different maize crosses kernel sizes. On the other hand, a population of kernels of a maize cross (32x19), as homogenous population was studied by taking sample sizes of  25, 50, 100, 200, 300, 500, and 1000 k. Depending on Chi-square test, sample sizes of 50 and 100 k gave good estimate to the mean value of kernel weight. A previous research by Elsahookie et al (6) studied the best combination of sample sizes and number of replicates to estimate mean value of plant seed yield of maize and sunflower. They found that the best recommended combination was using sample size of 10 plants with 4 replicates. Hertzog (10) reported that sample size should be over 40, while Julious (12) recommended sample size of 12 per group. More studies on other objects of sample sizes are really required.