甜高粱株高性状影响因素分析方法比较

Comparison for influence factors of plant height in sweet sorghum by different methods

为检测甜高粱株高影响因素,调查30份甜高粱种质20个农艺性状,分别应用简单相关、偏相关、通径分析和逐步回归通径分析方法检测。结果表明,简单相关检测到节数、节间长、秆重、叶长与株高呈极显著正相关,相关系数分别为:0.610、0.760、0.720、0.547;穗粒重、穗柄长、锤度与株高呈显著正相关,相关系数分别为:0.436、0.450、0.440;偏相关检测到节间长、节数与株高呈极显著偏相关,偏相关系数分别为0.981和0.951;基于决策系数通径分析检测到节间长、节数、叶长、锤度、穗粒重和生育期对株高有显著影响;其中节间长、节数和叶长对株高有促进作用,决策系数分别为0.579、0.367和0.043;锤度、穗粒重和生育期对株高有抑制作用,其决策系数分别为-0.073、-0.060和-0.052;逐步回归通径分析检测到8个影响株高的性状,按照决策系数排序为:节间长>节数>叶长>锤度>生育期>叶宽>籽粒宽>穗长。逐步回归通径系数显著性检验表明,在α=0.05水平,节间长与节数之间及与其他性状间差异显著;不同分析方法均能检测到节间长和节数两个主要因素,而对于锤度、生育期、叶长、叶宽、籽粒宽、穗粒重等次要因素检测结果不同。基于决策系数的通径分析方法与逐步回归通径分析方法的结果最为接近。

In order to detect the influence factors of plant height of sweet sorghum, twenty agronomic traits of thirty sweet sorghum germplasms were investigated and simple correlation, partial correlation and path analysis and stepwise regression methods were used for testing in this study, respectively. The results were as follows： Simple correlation analysis detected that number of internodes, intemodes length, stem weight, leaf length and plant height were extremely positive correlated with the correlation coefficient 0.610, 0.760, 0.720, 0.547,respectively. Spike grain weight, pedicel length, brix and plant height were significantly positive correlated and the correlation coefficient were 0.436, 0.450 and 0.440, respectively; Partial correlation detected that intemodes length and number of internodes were significantly correlated with plant height and the partial correlation coefficients reached 0.981 and 0.951; Path analysis based on decisioncoefficient detected intemodes length, number of intemodes, leaf length, brix, spike grain weight and growth period had significant effects on plant height. Intemodes length, number of intemodes and leaf length promoted plant height and decision coefficient were 0.579, 0.367 and 0.043, respectively; Brix, spike grain weight and growth period had inhibitory effects on plant height and its decision coefficients were -0.073, -0.060 and-0.052, respectively. Stepwise regression analysis detected eight traits affected plant height, with decision coefficient ranking order： intemodes length 〉 number of intemodes 〉 leaf length 〉 brix 〉 growth period 〉leaf width 〉 grain width 〉 spike length. Significance test for stepwise regression coefficient showed that there were significant differences between intemodes length and intemodes number and other traits at 0.05 level. Two main factors： intemodes length and number of intemodes were detected by different methods, while test results of secondary factors such as bdx, growth period, leaf length, leaf width, grain width and spike grain weight were not the same. Test results between path analysis based on decision coefficient and stepwise regression were most close.