Improving the selection method of repeated orthogonal test in the agricultural science research

The repeated orthogonal trial is an important method in the agricultural science research, and is widely used in multilevel regression collocation optimization and important factor analysis of agricultural disciplines due to its advantages of typicality and comprehensive comparability. However, the great deviation in optimization analysis was found in the orthogonal trial of significant interaction. Taking multifactors orthogonal design of nonquantitative and mixedquantitative as an example, This study developed the higher polynomial regression model to improve the orthogonal trial optimization method, and this method not only solved the possible interaction among factors and the disturbance of the fixed level limiting for excellent selection, but also explained accurately the model deviation and disclosed the relationship between the research factors and trial indicators.