Sample size and modeling of plant variability using precision statistics in soybean counting traits.

Counting traits are commonly measured in soybean research, however, the number of plants to be sampled per experimental unit needs to be standardized for estimating precision statistics from these traits, considering experimental designs; The current study evaluated the impact of plant variability for selected soybean agronomic traits; established the ideal sample size to estimate experimental precision statistics; and modeled the response of plant variability per experimental unit to forecast experimental precision statistics; Trials were conducted with 30 genotypes in highlands and 20 in lowland areas of Brazil. A complete randomized block design with three repetitions was used, measuring four counting traits in 20 plants per experimental unit, which represents a total of 9000 plants. From these experiments, trials with sampling scenarios from 1 to 100 plants per experimental unit were simulated; Results showed an exponential decreasing response of the 95% confidence intervals as a function of the increase in sample size per experimental unit for all precision statistics, stabilizing when ≥ 21 plants per experimental unit were sampled. Twenty-one plants per experimental unit were considered representative for the soybean traits here studied. The response of the 95% confidence interval width to sample size enabled the creation of good-quality forecasting formulas and equations to predict experimental precision in experiments with soybean that evaluate these traits; The sample size recommendations made in this work can be applied in similar experimentations. [Display omitted] • Twenty-one plants per experimental unit were enough for soybean counting traits. • Precision statistics were predicted from subtropical-climate soybean counting traits. • Forecasting precision improved after reaching 21 plants per experimental unit. [ABSTRACT FROM AUTHOR]