OPTIMAL SAMPLE SIZE FOR STATISTICAL ANALYSIS OF WINTER WHEAT QUANTITATIVE TRAITS
In the planning phase of every research particular attention should be dedicated to estimation of optimal sample size, aiming to obtain more precise and objective results of statistical analysis. The aim of this paper was to estimate optimal sample size of wheat yield components (plant height, spike length, number of spikelets per spike, number of grains per spike, weight of grains per spike and 1000 grains weight) for determination of statistically significant differences between two
... tween two treatments with significance level of 5% and statistical power of 0.8. Arithmetic means and standard deviations of winter wheat quantitative traits, statistical power, significance level α and effect size were used for calculation of optimal sample sizes. All calculations are made on data obtained from the experiment set up according to completely randomized design with two treatments in four repetitions. Optimal sample sizes for identification of ten percent difference between treatments are calculated for all traits. Retrospective power analysis was conducted for original data set of n=440 and optimal sample sizes are estimated. Obtained results showed that every examined yield component requires unique sample size. In terms of pertaining standard deviation and minimum expected difference that should be detected as statistically significant. For analyzing number of spikelets per spike sufficient sample size is 38 for effect size of 10% at significance level of 5%. Number of grains per spike and grain weight per spike (g) are the most variable traits and according to that, they have the largest estimated sample sizes (n = 440 and n = 436 respectively). Optimal sample size is mostly dependent on variability of examined trait and effect size.