Fractional factorial split-plot row column designs

Stapleton, Robert David (2005) Fractional factorial split-plot row column designs. University of Southampton, Doctoral Thesis.

Abstract

Factorial designs are considered in which the experimental units are arranged in blocks (wholeplots) and, within each wholeplot, in a row column array.  Suppose that, due to some practical or economic reasons, some factors in the experiment are considered more difficult to change (wholeplot factors) than others (subplot factors).  A combination of the levels of the wholeplot factors is then applied to each wholeplot and, within each wholeplot, level combinations of the subplot factors are allocated, at random, to the experimental units.  To reduce the number of experimental runs in the design, a fraction of the total number of possible factor level combinations is performed.  In this thesis, such designs are called fractional factorial split-plot row column (FFSPRC) designs and are considered.  (In particular, we consider designs where each factor has two levels, a low and a high level.)

The structure of FFSPRC designs is described for 2-level factors and a tech­nique for constructing designs presented and illustrated.  A method for identifying non-isomorphic designs is presented and an algorithm is given for constructing an exhaustive set of non-isomorphic designs.  A catalogue of designs is presented for up to five wholeplot and five subplot factors, and involving up to 512 runs.  Three techniques for distinguishing between competing designs using aberration and esti­mation capacity are proposed and illustrated.  The ideas are also extended to the construction of designs where each factor has the same number of prime, or prime powered, levels and illustrations are given.

This record has no associated files available for download.

Contributors

Download statistics