Augmented RBD for 60 maize hybrids in early-stage testing
60 test hybrids, 4 checks, 3 blocks. Limited seed per hybrid. Adjusted means rank candidates and decide which 12 advance to the next season.
Vinod is screening 60 single-cross maize hybrids in early-stage testing. Seed quantity is barely enough for one plot per hybrid. Four commercial checks (CK1, CK2, CK3, CK4) have plenty of seed. The objective is to advance the top 12 hybrids into the next season's replicated trial.
Design choice
Augmented RBD with 3 blocks. Each block contains all 4 checks (replicated across blocks) plus 20 of the 60 test hybrids (non-overlapping). 4 + 20 = 24 plots per block, 72 plots total.
Data, in StatVeda format
block,entry,type,yield_t_ha B1,CK1,check,7.21 B1,CK2,check,7.85 B1,CK3,check,8.02 B1,CK4,check,7.61 B1,T01,test,8.41 B1,T02,test,7.62 ... (T03 to T20 in B1) B2,CK1,check,7.45 B2,CK2,check,8.12 B2,CK3,check,8.31 B2,CK4,check,7.92 B2,T21,test,8.71 ... (T22 to T40 in B2) B3,CK1,check,7.05 B3,CK2,check,7.71 B3,CK3,check,7.85 B3,CK4,check,7.41 B3,T41,test,8.55 ... (T42 to T60 in B3)
What he does in StatVeda
Open Experimental Designs category. Pick Augmented RBD. Paste the long-form CSV. Tell the tool which entries are checks (toggle the type column to check or test). Run.
Block effects table: B1 sits 0.18 t/ha below the check grand mean, B2 sits 0.32 above, B3 sits 0.14 below. The block effect is real but modest.
Adjusted means table: each test entry mean has been corrected by subtracting its block effect. T21 raw 8.71, adjusted 8.39. T01 raw 8.41, adjusted 8.59.
Three CD values reported at 5 percent. CD(check vs check) = 0.41 t/ha, CD(check vs test) = 0.62, CD(test vs test) = 0.79.
Best check is CK3 with adjusted mean 8.06. Twelve test hybrids exceed best check by more than CD(check vs test): T01, T15, T21, T29, T34, T41, T47, T49, T52, T55, T58, T60.
What he decides
Advance T01, T15, T21, T29, T34, T41, T47, T49, T52, T55, T58, T60 to the next-season replicated yield trial.
Drop test hybrids whose adjusted mean is below worst check (CK4-adj. 7.55).
Keep CK1, CK2, CK3, CK4 as checks in the next-season trial for continuity of comparison.
Vinod attaches the StatVeda PDF export to the breeding programme report. The 12-line shortlist is approved without further discussion because the CD logic and the block-effect adjustment are visible on the same page.
Why this design instead of a replicated trial
A fully-replicated 60-entry trial with three reps would need 180 plots per site. With one site of 72 plots Vinod fitted everything in one block of land, kept the trial layout uniform, and still got a defensible ranking. The cost was that test entries are unreplicated within site. He accepts that cost because at this stage he is screening to drop the bottom 80 percent, not measuring small differences between the top 20 percent.
What gets carried forward
The 12 advanced hybrids go into a multi-location replicated trial next season, three reps per site, four to six sites. At that stage AMMI biplots become the right tool to read GxE structure. The Augmented RBD did its job: it filtered the candidate pool from impractical (60) to manageable (12) without spending three seasons worth of seed.