How many cauliflower seedlings are necessary to estimate experimental precision statistics reliably?

The aims of this study were to determine the adequate sample size of cauliflower seedlings per experimental unit to estimate experimental precision statistics, and to forecast experimental precision statistics. From a reference experiment in which 20 seedlings per experimental unit were sampled and the number of leaves, shoot length, root length, and total length were measured, a bootstrapping procedure was applied with replacement, recreating experiments with sample sizes ranging from 1 to 100 seedlings per experimental unit, obtaining 10,000 resamples for each size. For each resample, 13 precision statistics were estimated, for which the 95% confidence interval widths were calculated, fitting these values to the sampling scenarios through traditional power models. The sample size was then determined through the maximum curvature point through the perpendicular distances' method for each precision statistic and trait, for which a modified power model was used for forecasting. Also the contribution of the precision statistics to sample size definition was verified via principal component analysis. The reference experiment showed high experimental precision, which varied according to the trait and precision statistic considered. An exponential decrease in the confidence interval width was observed as the sample number was increased, and the optimal sample size varied slightly among statistics and traits, where the maximum value found was 20 seedlings per experimental unit, however, considering the statistics that contributed the most to the total variation, 16 seedlings per experimental unit represent the sufficient sample size for estimating experimental precision statistics in experiments with cauliflower seedlings. Also, the forecasting models showed excellent fitting quality and can be used and adapted by researchers performing experiments on seedling production of cauliflower or other horticultural crops.