| Summary: | Biological and agricultural experimental research aiming to be scientific have to deal with permanent control and variability analysis of experimental units under study, primarily by knowing and defining the intervals of allowed unexplained variations that can be tolerated in the samples as representative estimates of the mean value of a population. Numerous results of these analyses in biological and agricultural sciences show this interval is determined by variation coefficients ranging from 5 % ≤ Vk≤ 30%. The issue of data variability in samples has to be seriously considered because when testing significance of differences between mean values of applied treatments with small variations (below 5%) unrealistically very large number of statistically significant differences occur, whereas with big variations (over 30%), indicative experimental differences are estimated as accidental, that is, they remain unobserved. At the same time, this means that variation coefficients below 5% and over 30% have to be additionally checked as variation coefficients below 5% in these studies show that the results are "too good" to be approved as accurate without checking, whilst variation coefficients over 30% point to systemic effect of unobserved factors on particular statistical units under study due to which those are not representative experimental results.
|
|---|