Analysis of residuals and adjustment in JRA

• Author(s): Oliveira, Amilcar ; Oliveira, Teresa ; Mexia, João Tiago
• Date: 2008
• Persistent ID: http://hdl.handle.net/10400.2/2058
• Origin: Repositório Aberto da Universidade Aberta
• Subject(s): Joint Regression Analysis; Ziz-zag algorithm; Residuals
• Document Type: Journal article
• Language: English
• Contributor(s): Repositório Aberto

Description

Joint Regression Analysis (JRA) is based in linear regression applied to yields, adjusting one linear regression per cultivar. The environmental indexes in JRA correspond to a non observable regressor which measures the productivity of the blocks in the field trials. Usually zig-zag algorithm is used in the adjustment. In this algorithm, minimizations for the regression coefficients alternate with those for the environmental indexes. The algorithm has performed very nicely but a general proof of convergence to the absolute minimum of the sum of squares of residues is still lucking. We now present a model for the residues that may be used to validate the adjustments carried out by the zig-zag algorithm.