科研之友
微信
新浪微博
Facebook
分享链接
摘要
Tolerances of the International Seed Testing Association regarding the dispersion of germination test replicate results are based on the theoretical variance of a binomial sample. Several publications over the last 60 years give evidence that the dispersion between replicate results is in practice smaller than this theoretical variance. This article investigates in detail, whether this phenomenon indeed exists and if yes, what are the causes. Simulations show that the theoretical reference formula used up to now for the quantification of dispersion and for the establishment of tolerances is correct if sampling is at random and the evaluation unbiased. Nevertheless, the four empirical replicate results of a germination test show on average, without rounding of their mean germination percentage, a statistically significant underdispersion. The laboratory where the test was conducted had the highest influence on this underdispersion within a given species. These findings indicate clearly that current procedures for the germination test, whereby the four replicates are evaluated by hand usually by the same person, cause a bias and this bias causes underdispersion. To avoid this and improve the correctness of the current tolerances, replicates of germination tests should be blinded and or tested by different seed analysts.
- 出版日期2016
