Abstract
Efforts were made to develop robust statistical models to evolve a crop-logging technique for identifying best yield indicators of banana (cv. Robusta), across its different growth stages. Optimum values of the identified variables were also worked out based on various biometrical characters of 200 plants observed at a farmer's field. Statistical models developed and validated showed that at 70 days after planting (DAP), number of leaves and plant girth [Coefficient of Determination (R2) 88%] with optimum values as 8 leaves and 15.07 cm were the best yield indicators. While at 126 DAP, plant girth and number of leaves (R2 = 89%) with optimum values as 34.5 cm and 12 leaves and at 185 DAP, plant height and leaf length (R2 = 92%) with optimum values as 138.9 cm and 127.3 cm were significant crop-logging parameters. Further, results showed that at 250 DAP, plant height and plant girth (R2 = 90%) with optimum values as 159.21 cm and 67.8 cm and at 315 DAP, leaf breadth and leaf length (R2 = 81%) with optimum values as 67.2 cm and 164.1 cm were the significant yield predictors. Finally, during harvest stage, i.e., at 375 DAP, number of fingers per bunch, and number of hands per bunch (R2 = 99%) with optimum values as 26 fingers hand-1 and 13 hands bunch-1 were the best indicators of crop yield. Identified models were also made robust by removing the effect of multicollinearity among biometrical characters. Before making a final decision about the model adequacy, randomness assumption about the error term was statistically tested.
