Incomplete block designs for genetic testing: some practical considerations

Full-sib progeny and clonal trials to evaluate family and clone performance were simulated on a test site with various patterns of environmental variation to examine the effectiveness of randomized incomplete block designs (ICBs) under different design parameters (constant block size, variable block size, block shape and orientation, and family size). Simulations showed that blocks of fixed size 5-20 were effective in removing most site variation in a test with 120 full-sib families of three seedlings per family. Slight variations in block size did not affect the precision of mean estimations much when blocks were considered random. Square blocks were more effective than either column or row blocks in removing both patchy variation and gradients in one direction. With blocks of fixed size 10, increasing family size from 3 to 12 (at the expense of family number, which dropped from 120 to 30) reduced the average variance of family mean contrasts by 75%; most of this reduction (>50%) was achieved at 6 seedlings per family. The differences in precision of mean estimation between full-sib progeny and clonal trials were small with respect to the design parameters and patterns of site variation examined.