PROCEDURE OF THE GENERALIZED LINEAR MODEL FOR THE ANALYSIS OF AGRICULTURAL RESEARCHE RESULTS

Abstract

Orel, e-mail: melnik.anat202@yandex.ru, shumetov@list.ru The main stages of agricultural research results` modeling using the procedure of generalized linear model are considered; its advantages in the analysis of field experiments are shown. Concrete examples of using the procedure for assessing the statistical significance of the factors` influence of agrotechnical experiments, building confidence areas of sample parameters and testing hypotheses are given. It is shown that the model adequately reflecting the effect of precursors (corn for green mass, peas for grain and barley) and fertilizer rates (2 and 4 centners / ha of nitrogen-phosphorus-potassium fertilizer) on the yield of winter wheat of variety Moscovskaya 39 is a two-factor linear model of analysis of variance, and predecessor is more powerful than the norm of mineral fertilizers. ls method of the model parameters, the confidence intervals of the effects of the precursors and fertilizer rates were obtained. It has been proven that the precursor peas and a nitrogen concentration of 4 centners / ha are optimal for increasing the yield of winter wheat. Homogeneous groups of predecessors were formed according to Tukey’s multiple comparison criterion, with the predecessor «peas» forming an independent subgroup providing greater yields of winter wheat, while barley and corn are included in the common subgroup of predecessors providing lower yields. Two-factor linear models of analysis of variance were obtained, which also adequately reflect the influence of the precursors and fertilizer norms on the grain quality indicators of winter wheat variety Moscovskaya 39 – the content of crude protein and gluten. It is proved that, as well as for yields, the predecessor peas and the norm of nitrogen-phosphorus-potassium fertilizer are 4 t / ha. It is proposed to visually assess the quality of modeling by comparing the diagrams of dependence of productivity indicators on the factors` levels, built on the basis of actual and calculated data. A significant advantage of using the procedure of generalized linear model for analyzing the results of field experiments is the possibility of modeling according to average data, in the absence of information on replications, which makes it possible to build