Balanced block designs robust against the loss of a single observation (Q1118953)

An experiment based on a balanced block design (BBD) is usually no longer balanced if one or more observations are unavailable. This paper investigates the problem of finding a BBD which is most robust against the lack of balance induced by the loss of a single observation. Competing BBDs have the same number of treatments and observations, but different block sizes. Robustness of a BBD is measured through the non- zero eigenvalues of the C-matrix of the design which is missing the observation, either with the variability of those eigenvalues or with an A-optimality efficiency factor. Guidelines for finding the most robust design are given.