Generation of experimental designs with CycDesigN

The main objective of an experimental design is to evaluate and compare the response of several conditions or circumstances of interest. These situations are usually recognized as treatments and they are chosen for the evaluation of a research hypothesis of interest. These treatments are applied to several experimental units to control for variation. This variation, also called random or background noise, is inherent to any experiment and it can be caused by several factors, such as natural or biological variation, measurement errors, and interactions with uncontrolled factors.

The basic elements of design of experiments are: randomization, replication and blocking. With randomization, we ensure that treatments are assigned without bias to the different experimental units. Replication allows us to increase the precision of the experiment by observing the effect of a treatment several times, and by separating background noise from the real treatment effect. Finally, blocking provides control, up to a certain degree, of environmental variation by grouping similar units together.

CycDesigN is a software designed to generate optimal or near-optimal experimental designs with one- or two-dimensional blocking structures for a range of design parameters and design types. The package includes options to produce resolvable and not resolvable incomplete block and row-column designs. It also allows incorporation of several treatment structures such as single, nested or factorial. Additionally, latinized designs for resolvable designs with neighbouring full replicates, partially replicated designs and cross-over designs can be produced.

CycDesigN also includes a facility to prepare GenStat (and SAS) code for analysis of core design types.