Sample size determination in horticultural research: An empirical approach

- Virtual samples were created through resampling the measurements of a single sample.
- Sample size N=a(CV/Δ)2+b,a and b known for a given confidence probability.
- Pilot-sample size n=3.52(98.816−y)/y,y − predefined pilot-sample dispersion.
- The equations are universal provided the sample estimates are normally distributed.

Proper sample size determination is important for the representativeness of the research and the validity of the drawn conclusions. It is usually done iteratively because, under normal distribution, the number of trials N has a complex relationship with the mean X¯, the standard deviation S of a sample, the chosen confidence probability p, the predefined standard error Δ and the Student's criterion t. The present paper offers a simplified sample size calculation based on the relationship between N and the ratio (CV/Δ)2, where CV is the variation coefficient of a pilot sample. It was established that the sample size might be calculated with a highly satisfying precision from the formula: N=a(CV/Δ)2+b, the parameters a and b being dependent on p. Furthermore, it was concluded that in all experiments under normal distribution, irrespective of the studied subject, the size of the pilot sample might be evaluated through the nonlinear nonparametric equation: n=3.52(98.816−y)/y,where y is a predefined value of CV of CV of the series of measurements.