Abstract

The aim of this study was to show that repeatability is statistical value related with the experimental precision in the analysis of experiments. A database of the rice project of the Agricultural Research Institute of Panama was used. 379 individual experiments were analyzed, grouped in 41 trials, planted in the main rice growing-areas in Panamá, between the years of 2000 and 2014. The variance components were calculated through the individual and combined REML analysis, statistical calculations were performed, like the variation coefficient (VC), minimum significant difference (MSD), repeatability (H), range, determination coef cient (R2) and the MSD/ range ratio. To determine the appropriate repeatability values to incorporate the combined analysis and improve the experimental precision was done several times throughout the analysis. The individual and combined analysis showed a link between MSD/range ratio and repeatability. When discarding the sites with H<0, the precision was raised by 60% with trial discards under 0,20, 0,30, 0,40 y 0,50, and the precision levels were 38, 67, 56 y 50%, respectively. A convenience in including sites with H>0,20 was confirmed, and discarding those with lower H. There was a lineal link between the number of locations and repeatability, an increase in locations and repeatability, improving the precision. The link between number of locations and MSD/ range was inverse, when raising the number of locations the proportion in MSD/range was lowered, increasing the experimental precision. 

Keywords: precision statistics, agricultural research, experimental efficiency, variation coefficient.