Efficient generation of optimal or near-optimal experimental designs



Five general classifications of designs can be constructed.

CycDesigNGet a Quote
Download CycDesigNDownload CycDesigN


    CycDesigN is a computer package for generating optimal or near-optimal experimental designs. CycDesigN 7.0 is the latest version of the package. Five different types of designs are available, namely resolvable, non-resolvable, multi-location partially replicated (p-rep), single location p-rep and crossover designs. Apart from the last type, the designs can be arranged in blocks or in rows and columns.

    Resolvable designs have the treatments arranged in groups of blocks or rows and columns such that each treatment occurs once in each set. They can be latinized, t-latinized or partially latinized. The designs are needed when there is a requirement to allow for possible differences between blocks or rows and columns within replicates (sets). It is often useful in practice to perform the experiment one replicate at a time, and accuracy will be increased if the experimental material can be arranged so that the replicates are relatively homogeneous.

    Multi-location p-rep designs have test treatments appearing either zero, once or twice in each location, such that each of these treatments occur an equal number of times in the design. Standard treatments can also be incorporated into the designs. Both design types are widely used by experimenters involved in field trials.

    If the design cannot be resolved into replicate groups then it is a non-resolvable design. Again the design can be arranged in blocks or rows and columns. It is important that the experimental material used is relatively homogeneous so that the number of blocks, or rows or columns is usually fairly small. Hence, they are most suitable for smaller experiments in, for example, glasshouses or laboratories, including plant phenotyping experiments.

    A new feature of CycDesigN 7.0 is the introduction of the new class of single location p-rep designs, as alternatives to unreplicated designs. Traditionally the latter have been used in early generation plant breeding to enable the selection of material from a large number of test treatments (genotypes) with limited quantities of seed. A number of standard treatments (controls) are typically included for comparative purposes and to provide an estimate of error. More recently it has become popular to duplicate a proportion of the test treatments, since this can improve design efficiency for the same experimental area. However it is usual for single location p-rep designs to still include standard treatments but the number of these is likely to be substantially reduced.

    The final type of designs is crossover designs; they are sometimes called change-over or carry-over designs. They are used when sequences of treatments are applied to several subjects over a number of time periods. The direct effect of the treatment applied in the current period and the carry-over effects of the treatments applied in one or more previous periods can be modelled in various ways, although in practice the additive model is more often employed. Crossover designs are widely used in clinical trials, as well as in taste, psychological and psycho-physical experiments.

    Another important new feature of CycDesigN 7.0 is the development of new algorithms for the construction of spatial designs. Available in this version are spatial enhancements of resolvable block and row-column designs, and of non-resolvable and multi-location and single location row-column designs. The aim is to achieve a good spatial separation of the treatments in the designs, using neighbour balance and evenness of distribution (NB & ED) principles.

    CycAnalysis is a module in CycDesigN that allows output from a CycDesigN session to be tailored in a form ready for analysis. For instance, a spreadsheet of the blocking and treatment structures is automatically generated. Among other features, it is possible to rename treatment factors and labels as well as blocking and plot factors. CycAnalysis also allows you to generate Genstat and SAS code for the analysis of most designs, including the new single location p-rep designs and all spatial designs.

    CycDesigN 7.0 provides the most comprehensive design generation package yet available for experimenters. In addition there is a self-contained and updated Help facility covering all of the extensive features available in the package. The authors are leading researchers in experimental design and, as a result, the algorithms incorporated in CycDesigN 7.0 are the most recent developments in the construction of experimental designs.

    New features:

    • New and improved licensing system.
    • A substantial revision of the algorithms used to generate resolvable, non-resolvable and partially replicated block designs. The algorithm now seamlessly constructs designs taking advantage of the speed of cyclic and alpha design construction, and automatically switches to a more general algorithm to search for more efficient designs.
    • The construction of non-resolvable designs with unequal treatment replication.
    • Spatial enhancement of both equal and unequal replicate non-resolvable designs.
    • The construction of partially replicated row-column designs.
    • The incorporation of standard treatments into the generation of efficient partially replicated block and row-column designs.
    • Genstat or SAS code for block and row-column partially replicated designs, with or without standard treatments.
    • An updated and improved Help facility that is now more self-contained.


    For the generation of optimal or near-optimal experimental designs.