Resumen
Plant loss occasionally occurs in field experiments with coffee crops in experimental plots. In breeding programs, such loss can be harmful, especially when the corresponding statistical analyses are not consistent with the experimentally generated data. Herein, we analyzed productivity data to determine whether the compensatory effect occurs in coffee crops, evaluated the need to correct experimental failures, and identified the best stand correction method. We used productivity data from six harvests of eleven experiments with Coffea arabica plants. The experiments were implemented in a randomized block design, with four replications and six plants per plot. The following stand correction methods were evaluated: rule of three; Zuber; Vencovsky and Cruz covariance of the average or ideal stands; and Cruz, and the data were compared without correction adjustments. The most adequate correction methods were selected based on genetic variance, selective accuracy, and progeny ordering. The compensatory effect was evident from the analyzed data, with stand correction being evidenced as beneficial in progeny competition experiments. The best results were obtained for the covariance methods using average or ideal stands, followed by the method proposed by Cruz. The rule of three and Zuber method exhibited unsatisfactory results and are not recommended for stand correction in progeny competition experiments with coffee crops.