大豆区域试验中的误差、基因型×环境互作及试验设计的研究

THE RELATION OF ERROR VARIANCE, GENOTYPE ×ENVIRONMENT INTERACTION IN THE REGIONAL TEST OF SOYBEAN TO EXPERIMENTAL DESIGN

摘要:

用方差分析法研究了1981～1985年吉林省大豆联合区域试验中95个点次的产量试验资料,计算了各点次的机误方差和机误变异系数及基因型×环境互作方差组分等参数。计算了检验不同水平的品种间差异需要的总重复数,研究了各种互作条件下同一环境内的区组数和环境数对试验准确性的影响及两者之间的相互关系,讨论了区域试验的合理设计问题。各熟期组平均机误均方为1.325,平均变异范围为0.606～1.957。各熟期组的平均机误变异系数为11.93％。所研究的试点中约有57.7％的点次机误变异系数小于10％。平均基因型×环境互作方差组分约是机误方差的0.112倍,其大小变化于0.000～1.553之间。假定无基因型×环境互作,当α=0.05,β=0.50时,如试验的机误变异系数为10％,要检验出两个品种间10,20,30％的差异分别需要10,3和2欢重复。当有基因型×环境互作时,增加试验地点内的重复数或试验环境数均可有效地降低试误差。在试验点内的重复数一定时,试验需要的环境数随互作的增大而增多。

Abstract:

Yield data of the regional uniform soybean tests of 95 environments in Jilin province from 1981 to 1985 Were studied with a variance analysis method. The error variances, coefficients of variation and the variance components of G×E interaction were calculated for each environment or maturing groups. The total number of replications for detecting various levels of differences between two strains were also calculated. The number of blocks within an environment and the number 0f environment effecting on the accuracy of a given experiment were examined under different G×E interaction levels. some problems concerning the design of regional test in soybean were discussed. The average error mean square over maturing groups was 1.325, and ranged between 0.606 and 1.957. The average coefficient of vaiation over maturing groups was about 11.93%. The coefficients of variation of about 57.7% out of the environment studied were smaller than 10%. The average G×E interaction variance component was 0.112 time of the error variance, and ranged between 0.000 and 1.553. Assuming no G×E interaction, an average of 10,3 and 2 replications was required to detect 10, 20 and 30% yield difference between two strains respeciively when the coefficient of variation was about 10% （α=0.05, β= 0.50）. The experimental error could be effectively reduced by increasing the number of replications within an environment, or the number of environments when G×E interaction was present. The numer of environments required increased with the magnitude of G×E interaction when the number of replications within environment was constant.