| Abstrakt: |
Data from different multi-environmental trails (MET) were analysed, including different number of varieties, number of locations and different research periods. The first experiment (24 PhD) included 24 wheat varieties that were studied in five locations of the country over a period of four years (2009-2012). The second field experiment (40 ABC) consists of 40 new advanced wheat lines and cultivars, which were studied in three locations over a three-year period (2017-2019). The grain yield datasets from the two experiments were used to make a direct comparison of various statistical parameters to assess the genotype stability against the background of significant growing conditions. The study involves the use of several statistical packages that are specialized for this purpose. Based on the ranking assessment of the values of each statistical parameter, a critical analysis was made of its relationship with the yield, for each dataset separately. For this purpose, the possibilities of correlation, principal component and cluster analyses were used. Parameters for which information differs between datasets or between statistical packages are removed from the analysis list. The final set of 31 parameters was analysed according to the set goal, after a statistically justified possibility to merge the two datasets. Most of the rank parameters do not show correlation with grain yield. The units are the parameters, the correlation of which is either positive (Pi, Ysi, TOP, λ) or, respectively, negative (DJi, NP(1), CVi]). The analysis of the data through different statistical approaches shows that the parameters correspond to the dynamic concept of stability assessment. Only one of the parameters (θi) is related to static stability assessment. In the presence of many more effective than it, it should not be applied because it is an exception from the analysed group. The groups of parameters of the regression coefficient (bi), the deviation from the regression line (s2di), ecovalence (W2i) and the stability variance (σ²i), give objective information about the behaviour of the variety in environmental conditions and it is not influenced by software. Some of the non-parametric [S(i) NP(i)] assessment methods provide diametrically opposed information for stability because of differences arising from either the dataset or the software used. Suitable for stability assessment are non-parametric approaches - [S(1) and S(2)], which is fully confirmed by the three software packages. Each of the used software packages contains a set of parameters, the application of which as a set gives correct information about all aspects of the wheat stability [ABSTRACT FROM AUTHOR] |
