摘要
采用随机区组设计进行品种区域试验,在多重比较时,对照品种与参试品种的差异显著性可用最小显著差数浊(L.S.D)测验。由L.S.D可推导出CV≤D·√n/2/tα。由此可推算出一定概率平准保证下的精确度要求。按重复数3、4、5、6、推算了一定平均差数百分率(D=(B-A)/x×100%)下,5%概率平准容许的最大变异系数,列于表1。对江苏省近几年来小麦、大麦区试的历史资料进行分析,其平均CV约为6%,这样水平的CV仅适于测验D>20%的试验。而这样大小的差异,在大麦和小麦近年来的区域试验中,分别仅占17%和30%以下。结果表明,各试验点现在的CV水平过高,必须予以降低。对改进区域试验精确度的途径进行了讨论。增加重复次数也许是最有效的途径之一。
For the purpose of obtaining a probabilityαof a significant difference between a pair of cultivars and based on the knowledge of a good estimate of the standard error from the past experience,the number or replications(n)suggested to be used can he calculated with the formula of n≥(σ/δ)^(2)·t^(2)_(0.05).The value of n depends only on theσ/δand is independent of the units used forσandδprovided they are the same for both.Hence,if they are standardized by the overall mean,the formula can be rewritten as:CV≤D√n/2/t_(0.06).Using the formula as an equation for CV and fixing n=3,4,5 and 6,the required CV%for testing various differences at the tis shown in table 1.Historical data from wheat and barley cultivar regional tests(1981-1985)in the Jiangsu Province were used to find out the distributional pattern of participating locations with respect to the coefficient of variation(CV)and D,the difference between the best check cultivar and the best test cultivar expressed as percentage of the mean at each location.The average CV in these regional tests is about 6%,which is adequate for detecting D values greater than 20%only.However,differences of this magnitude occur in no more than about 17%and 30%of recent trials for harley and wheat,respectively.These results suggest that the present levels of CV within location are too high and need to be reduced.Ways to improve precision of these regional tests are discussed.Increasing the number of replications is one of the most effective ways.
作者
俞世蓉
吴兆苏
Yu Shirong;Wu Zhaosu(Wheat Breeding Institute,Nanjing Agrieultural University,Nanjing,Jiangsu Province)
出处
《中国农业科学》
1987年第6期10-16,共7页
Scientia Agricultura Sinica
