Trial Designs

Alpha-Lattice

1.0

A kind of resolvable incomplete block design, wherein, not all treatment levels are tested in the same block, and it is a resolvable design because blocks are arranged and group together to form complete replications.

Design parameters: number of occurrences (trials), number of replicates, number of blocks.

Layout parameter: number of rows per block, number of rows per replicate, total number of field rows, plot ordering (serpentine or not)

Usage: When the number of treatments being tested (entries, genotypes) is very large, it is difficult to have a uniform field that supports this large number of plots, so we build small blocks (incomplete blocks) within the replication. Block size need not be large even when the number of treatments is large. More flexible than lattice designs (which require that number of treatments should be a perfect square and block size or number of plots in each block is equal to the square root of the number of treatments).

Alpha-Lattice | CIMMYT

2.0

Same as Alpha-Lattice.

Differences:

• In this one the user selects the size of the blocks, and not the number of blocks.

• Layout parameters:

  • Number of plots until field divisor (wrap the serpentine).

  • Serpentine options (vertical or horizontal)

Alpha-Lattice | WHEAT

2.0

Same as Alpha-Lattice | CIMMYT.

Differences:

• Add randomization degree options (randomize first replicate and usage of same randomization for all occurrences)

RCBD

1.0

Randomized Complete Block Design. Is the standard design for agricultural experiments, the replication is arranged in group, creating a local control factor that is called “Block”. Each block contains one plot of each entry, there are no theoretical limit for the number of blocks. The entries are randomized within the blocks. In this design block and replication are the same entity.

Design parameters: number of occurrences (trials) and number of replicates (blocks)

Layout parameters: number of field rows, rows per replication (block) and plot ordering (serpentine or not).

Usage: When the experimental units are not homogeneous and it is possible to group the experimental units into blocks such that experimental units within each block are more homogeneous than those between blocks. It is not recommended when the number of entries are too large, being impossible to have field with a little variation even within the blocks.

RCBD | CIMMYT

2.0

Same as RCBD

Differences:

• Layout parameters:

  • Number of plots until field divisor (wrap the serpentine)

  • Serpentine options (vertical or horizontal)

Row-Column

1.0

Another kind of resolvable incomplete block design, like Alpha-Lattice, but with better field variation since this design allows blocking in two directions that are perpentidular to each other.

Design parameters: number of occurrences (trials), number of replicates, number of row blocks.

Layout parameter: total number of field rows, plot ordering (serpentine or not)

Usage: When the number of treatments being tested (entries, genotypes) is very large, it is difficult to have a uniform field that supports this large number of plots, so we build small blocks (incomplete blocks) within the replication. This design is appropriate if there are two sources of variation and the dimension of the field per replicate is squarish as possible.

Augmented RCB

1.0

Another type of incomplete block design wherein some of the treatments (entries, genotypes) are replicated while the the other treatments are unreplicated. Replicated treatment are referred as check while the unreplicated are referred as the test. The replicated treatment follows the randomization of the RCBD (all treatments are present in each block) while the unreplicated treatment are distributed and embedded within each block.

Design parameters: number of occurrences (trials), number of replicates, number of blocks.

Layout parameter: total number of field rows, plot ordering (serpentine or not)

Usage: This design is often used in the early generations, when seed is limited or resources are limited, want to evaluate as many genotypes as possible, difficult to maintain homogeneous blocks when comparing so many genotypes