TY - JOUR
T1 - Construction of crossover designs with correlated errors
AU - Williams, E. R.
AU - John, J. A.
PY - 2007/3
Y1 - 2007/3
N2 - Crossover experiments are widely used, particularly where a sequence of treatments is given to subjects. Correlations between observations on the same subject are therefore likely and should be considered in both the design and analysis of crossover experiments. This paper presents an algorithm for the generation of efficient crossover designs with autoregressive and linear variance structures. The algorithm has been implemented as a module in the experimental design generation package CycDesigN (Release 3.0; CycSoftware, Hamilton, New Zealand). Output from the algorithm is compared with earlier work. Some results are given from the analysis of a crossover experiment assuming correlated errors.
AB - Crossover experiments are widely used, particularly where a sequence of treatments is given to subjects. Correlations between observations on the same subject are therefore likely and should be considered in both the design and analysis of crossover experiments. This paper presents an algorithm for the generation of efficient crossover designs with autoregressive and linear variance structures. The algorithm has been implemented as a module in the experimental design generation package CycDesigN (Release 3.0; CycSoftware, Hamilton, New Zealand). Output from the algorithm is compared with earlier work. Some results are given from the analysis of a crossover experiment assuming correlated errors.
KW - Autoregressive variance structure
KW - Crossover experiments
KW - CycDesigN
KW - Experimental design and analysis
KW - Linear variance structure
UR - http://www.scopus.com/inward/record.url?scp=33846653107&partnerID=8YFLogxK
U2 - 10.1111/j.1467-842X.2006.00463.x
DO - 10.1111/j.1467-842X.2006.00463.x
M3 - Article
SN - 1369-1473
VL - 49
SP - 61
EP - 68
JO - Australian and New Zealand Journal of Statistics
JF - Australian and New Zealand Journal of Statistics
IS - 1
ER -