Aspects of the Design and Interpretation of Competition (Interference) Experiments

Abstract

The design and analysis of competition experiments should be based on an unambiguous objective. Recent criticisms of particular designs have been made without reference to objectives and may be misleading. Three common designs are discussed: additive, replacement series, and response surface. Additive designs are well suited to agronomic objectives; replacement series are useful for comparing pairs of species; response surface designs can be used for most objectives but may be unnecessarily complex. The published criticisms of additive and replacement series designs are argued to be acceptable limitations within the bounds of the objectives for which they are used. Concerns about these designs confounding density and proportion are irrelevant to the objectives for which they are most suited. The continued use of multiple comparison tests is argued to be illogical. Regression approaches to analysis are more relevant, many non-linear equations are now easy to fit to data and these can be used without the need for linearization. However, there are various pitfalls not adequately reported to date. In particular, error structures need to be checked carefully and over-elaborate equations should be avoided.