The suitability was assessed of various designs for field
experiments investigating plant diseases
caused by airborne pathogens that can be subject to interplot interference.
Use of a model to describe
such interference showed that the treatments with the most dissimilar
effects on controlling the
disease should be allocated to experimental plots furthest apart in each
block, in order to minimize
the interplot interference within a block. When using large square plots,
rectangular blocks were more
efficient than square blocks in minimizing treatment-comparison biases
due
to interference between
neighbours. For rectangular blocks with the square plots side by side,
less
biased treatment
comparisons were obtained from designs with complete blocks than from
designs with incomplete
blocks, especially when larger numbers of treatments were included in the
experiment. However,
when interplot variance is taken into account, incomplete blocks may give
better treatment
comparisons. Similarly, unbalanced designs composed only of incomplete
blocks
that yield less biased
treatment comparisons may be better than balanced incomplete block designs
when interplot
variance is low. For high levels of variation, balanced incomplete block
designs may be more
appropriate, as increasing the precision of the treatment comparisons
becomes more important than reducing the bias.