Biplot analysis has been increasingly used in visual analysis of genotype-by-environment data and other types of two-way data. While many plant breeders and agricultural researchers are enthusiastic about the capacity of biplot analysis in helping them to understand their research data, some statisticians consider the use of biplots as a sidetrack to genotype-by-environment interaction analyses. Confusion also exists among statisticians on what is or is not a biplot. Admittedly, some users of biplot analysis are not always clear on how to select a proper type of biplot for a particular research objective and how to interpret a biplot correctly, accurately, and adequately. Some criticisms of biplot analysis may arise from incomplete understanding of the practitioners’ research problems as well as of the biplot methodology. In this review, I summarize the experiences and understanding in biplot analysis of genotype-by-environment data achieved during the last decade and discuss the following issues: (1) how to choose a proper biplot; (2) how to choose a proper GGE (genotype + genotype-by-environment interaction) biplot; (3) how to use the key functions of a GGE biplot for genotype evaluation, test-environment evaluation, and mega-environment delineation; (4) how to judge the adequacy of a 2-D biplot; and (5) how to test the statistical significance of a biplot pattern.

双标图分析越来越多地被用于直观分析农作物品种多点试验数据和其他类型的两向数据。这种方法深受植物育种家和农业研究人员的推崇, 认为它可以提高研究者理解和驾驭试验数据的能力；但也受到一些学者的批评, 认为它是统计分析方面的旁门左道。事实上，学术界对什么是双标图的认识尚存混乱。一些双标图的使用者并不总能正确地选择和解释双标图。一些双标图的批评者对双标图分析及其研究对象也缺乏深入了解。为使研究者对双标图分析有一个客观全面的认识, 本文就用双标图分析农作物品种多点试验中的几个问题进行阐述：(1) 如何针对特定的研究目的选择适当的双标图; (2) 如何选择适当的GGE双标图来分析多点试验数据; (3) 如何使用GGE双标图的不同功能形态进行品种评价、试验点评价和品种生态区划分; (4) 如何判断双标图是否充分表现试验数据中的规律; (5) 如何检验双标图显示的结果是否显著。

将原始数据减去各试点均值后形成的数据集中只含基因型主效G和基因型与环境互作效应GE, 合称GGE. 对GGE做单值分解, 以第一和第二主成分近似, 按第一和第二主成分值将所有品种和试点绘于同一平面图即形成GGE双标图. 以其分析我国春麦区10个试点20个品种淀粉糊化特性的峰值粘度, 结果表明铁春1号在大部分试点峰值粘度表现较好,

摘　要： 为了鉴定高产稳产玉米新品种以及为玉米高产栽培提供理论依据,通过基于R语言的GGE双标图法分析了10个玉米新组合在四川省甘孜州高海拔地区生态条件下的丰产、稳产性和适应性,同时分析了5个试验点对参试组合的辨别力和代表性。结果表明：‘康玉试0903’和CH327 2个组合高产且稳产;而SZ694较高产,但稳产性稍差,在海螺沟试验点有特殊适应性。综合各试验点的区分力和代表性,丹巴、九龙和海螺沟较好：九龙和海螺沟的区分能力强,丹巴试验点的代表性最佳。

Peanut planting density directly affects peanut yield and yield components, it is an effective way to adopt an optimal growing system for improving yield level. The paper presents an optimum density of peanut cultivation and a new method of data analysis. Five treatments (densities at 7.5×10 4 , 105×10 4 , 150×10 4 , 195×10 4 , and 240×10 4 holes per hectare) were designed with peanut cultivar Jihua 4. We investigated the agronomic traits, economic traits, yield components and pod yield in peanut, analyzed the differences between growing densities and every trait determined by SAS (Statistical Analysis System) and GGE (Genotype main effects plus Genotype×Environment interaction model). The results showed that there were inconsistent performances between traits and densities. The number of flowers per plant in the whole growth period, the number of pods per plant, pod yield per plant and pod yield per hectare were more sensitive to peanut planting density, but the main stem height, lateral branches length, shelling percentage, oil yield per hectare and protein yield per hectare were more stable. The number of flowers per plant in the whole growth period, the number of pods per plant, 100-pod weight and pod yield per plant decreased significantly with the increase of planting density, but pod yield per hectare increased. Pod yield produced by plants with the highest planting density were slightly greater than those with the lower planting density, while with planting density increasing, the increasing degree of pod yield per hectare decreased. Although the total pod yield increased with density, the higher yield achieved at the higher density did not give a higher economic profit which also affected by seed cost for planting. The result showed that the optimal planting density of Jihua 4 was from 105.0×10 4 to 150×10 4 holes per hectare based on maximal economic profit. GGEbiplot is a new and more realistic and effective method to analyze growth and yield components of high oil peanut cultivar.

花生群体密度直接影响花生产量及产量构成因素，系统研究不同种植密度下花生产量和性状差异，可为花生高产栽培提供理论依据。以高产高油花生冀花 4 号为材料，设置 5 个密度处理，分别为每公顷 7.5 、 10.5 、 15.0 、 19.5 和 24.0 万穴，探索种植密度对高产花生农艺性状、经济性状、产量形成因素和经济产量的影响，并用 GGE 双标图法对结果进行系统分析。结果表明，不同指标对种植密度的敏感程度表现不一，单株开花数、单株结果数、单株产量和荚果产量表现最敏感，而主茎高、侧枝长、出米率、籽仁含油量及蛋白质含量表现相对较稳定。随花生密度增大，单株开花数、单株结果数、百果重和单株产量显著降低，荚果产量则逐渐提高，但提高幅度逐渐降低。冀花 4 号种植密度每公顷 10.5~15.0 万穴时可获最大经济效益。 GGE 双标图为研究不同密度下花生生长状况和产量反应提供了更为直观有效的分析手段。

采用GGE双标图法对2008～2009年黑龙江省5个试验地点的9个大豆品系试验数据进行分析.结果表明:嫩00-2111属于高产、稳产性较好的品系,平均产量2 508.1 kg·km-1;庆农05-1071和庆农05-1028在大庆试点表现较好,其产量分别是3 513.8和3 434.8 kg·km-2,具有特殊适应性,在局部地区有推广价值.在5个试点中,杜蒙试点最好,既具有代表性又具有较强的鉴别力.

This paper reports analysis of a 5×4 diallel cross of B. napus L. using GGE biplot. The following patterns were visualized from the biplots. (1) Male sterile lines S2 and S3 had high general combining ability ( GCA ), and S3 and S1 had high specific combining ability ( SCA ). Restorer lines R2 and R3 had high GCA , while R2 and R4 had high SCA . The male sterile lines S5 and the restorer R2 were the best parents in terms of both discriminating ability and representativeness. The combinations with high SCA included S4×R4, S3×R1, S3×R2, S5×R3 and S1×R3. (2) Across the sterile lines, the GCA of pods per plant and seeds per pod was positively correlated with the GCA of yield, and the GCA of plant height, 1000-seed weight, and effective branch height was negatively correlated with the GCA of yield. Across the restorer lines, the GCA of pods per plant, pods of main raceme, primary branches, length of main raceme, and plant height were positively correlated with the GCA of yield, whereas 1000-seed weight and effective branch height were negatively correlated with the GCA of yield. Across the hybrids, pods per plant and seeds per pod were positively correlated with yield, while 1000-seed weight and effective branch height were negatively correlated with yield. These results suggest that although good male sterile lines and the restorer lines have different characteristics, more pods per plant, smaller 1000-seed weight, and lower effective branch height are desirable for both. The trait profiles should be complementary between parents to generate superior hybrids.

用GGE双标图法对甘蓝型油菜5×4双列杂交试验进行分析,显示了以下结果: (1) GCA 较高的亲本为不育系S 2 、S 3 和恢复系R 2 、R 3 ，而SCA较高的亲本是不育系S 3 、S 1 和恢复系R 2 、R 4 ，不育系S 5 和恢复系R 2 是区分能力和代表性最好的亲本，SCA较高的组合为S 4 ×R 4 、S 3 ×R 1 、S 2 ×R 2 、S 5 ×R 3 和S 1 ×R 3 ；(2）对不育系而言，单株角果数和角粒数的 GCA 与产量 GCA 呈正相关，株高、千粒重和有效分枝部位的 GCA 与产量 GCA 呈负相关。对恢复系而言，单株角果数、主花序角果数、一次分枝数、主花序长度和株高等性状的 GCA 与产量 GCA 呈正相关，呈负相关的有千粒重和有效分枝部位。对于杂交种而言，与产量呈正相关的性状为单株角果数和角粒数，呈负相关的为千粒重和有效分枝部位。表明虽然好的不育系和恢复系具有不同的特点，但相同之处为单株角果数较多，千粒重较小和有效分枝部位较低。同时在亲本选配中应注意优势性状互补。

GGE (genotype main effect plus genotype-environment interaction) biplot analysis, based on principal component analysis (PCA), is an effective method to assess the multi-environment trials (METs). It allows a full exploration of G×E interaction, variety yield stability, and test-site suitability. In the present study, we applied GGE biplot to evaluate the genotype effects and G×E interaction of sugarcane main characters and at the same time, analyze the stability of yield and quality traits, and representativeness and discrimination of testing site, with seven sugarcane varieties planted in seven different testing sites. The results showed that YZ05-51 had high yield per stalk and better yield stability than the other varieties, while FN39 and LC03-1137 had high yield per stalk but low stability. FN38, YG35, and ROC22 had high stalk numbers but low stability. FN39 and YZ05-51 had high cane yield and good stability, while FN38 and LC03-1137 had high cane yield but low stability. FN39 and GN02-70 had high sucrose content and good stability, while YZ05-51 and FN38 had high sucrose content but low stability. FN39 had higher sucrose content and better yield stability, while FN38, YZ05-51, and LC03-1137 had high sucrose yield but low stability. Ruili City and Lincang City in Yunnan Province had better representativeness and discrimination in yield per stalk, while Baosan City in Yunnan Province and Laibing City in Guangxi Province had better representativeness and discrimination in stalk numbers; Chongzuo City in Guangxi Province and Lincang City in Yunnan Province had better discrimination in cane yield, while Baise City and Liuzhou City in Guangxi Province had better discrimination in sucrose content; Baise City in Guangxi Province and Lincang City in Yunnan Province had better discrimination in sucrose yield. This study demonstrates a simple and effective method for analyzing data, and evaluating the yield stability and adaptability of sugarcane varieties involved in national regional trials and favorite test sites.

采用 GGE 双标图分析 7 个甘蔗参试品种在 7 个试点的产量和品质性状。结果表明 , 云蔗 05-51(YZ05-51) 为单茎产量高且稳定性强的品种 , 福农 39(FN39) 和柳城 03-1137(LC03-1137) 的单茎产量较高 , 但稳定性较差 ; 福农 38 (FN38) 、粤甘 35 (YG35) 和新台糖 22 (ROC22) 有效茎数较高 , 但稳定性较差 ; 福农 39 (FN39) 和云蔗 05-51 (YZ05-51) 为蔗茎产量较高且稳定性强的品种 , 福农 38 (FN38) 和柳城 03-1137 (LC03-1137) 蔗茎产量高 , 但稳定性较差 ; 福农 39 (FN39) 和赣南 02-70 (GN02-70) 为甘蔗蔗糖分较高且稳定性强的品种 , 云蔗 05-51 (YZ05-51) 和福农 38 (FN38) 甘蔗蔗糖分较高 , 但稳定性较差 ; 福农 39 (FN39) 为产糖量较高且稳定性较强的品种 , 福农 38(FN38) 、云蔗 05-51 (YZ05-51) 和柳城 03-1137 (LC03-1137) 的产糖量较高、但稳定性较差。云南瑞丽和云南临沧 2 个试点单茎产量的代表性和区分力较强 ; 云南保山和广西来宾 2 个试点有效茎数的代表性和区分力较强 ; 广西崇左和云南临沧 2 个试点蔗茎产量的区分力较强 ; 广西百色和柳州 2 个试点甘蔗蔗糖分的区分力较强 ; 广西百色和云南临沧 2 个试点产糖量的区分力较强。

用GGE双标图对2009-2010年度南方冬大麦品种区域试验进行分析,以图谱的形式阐述了参试品种在各试验点的表现、各品种的平均表现和稳定性、各品种的适应地区、各参试品种的相似性和环境的相似性及两品种之间的比较等问题,很好的解答了品种与环境互作的问题.双标图显示‘苏啤3号'是一个适应性很广的品种.产量相关性状的图谱分析表明:本年度试验中有效穗与产量相关性最高,是影响产量的制约因素.

摘　要：

研究利用AMMI模型对2007-2008年甘肃省马铃薯品种区域试验中的10个马铃薯品种、7个试点进行综合评价,分析10个马铃薯品种的稳定性及丰产性,以及7个试点的鉴别力.结果表明:各品种的稳定性大小为L0206-6＞ L0227-17＞ L0227-18＞天2008-8-2＞富薯3号＞CK＞ 34-126＞98-6-2＞35-62＞陇薯6号,其中L0206-6,98-6-2品种的稳定性最强,陇薯6号的稳定性较差；L0227-17,L0206-6,L0227-18以及天2008-8-2的产量最高,98-6-2品种的产量最低.各个试点的鉴别力也具有差异性,各试点的鉴别力顺序为天水＞会川＞安定＞静宁＞临夏＞秦王川＞宕昌,其中天水地区对品种的选择性最高,宕昌地区对品种的鉴别力低.AMMI模型中的主成分值,共解释总互作平方和的93％,比线性回归模型能更有效地分析基因与环境互作效应.