Table III
A comparison of statistical performances for experimental designs with sixteen species representing four functional groups along two axes of functional diversity. The factorial, complete and TropiDEP design (with categorical or continuous independent variables) are compared. Performances are relative to the reference factorial design. Corr(V1, V2) is the correlation between the estimators of the effects of variables 1 and 2. The complete design excludes the impossible combination of eight species in a single functional group. The TropiDEP design with continuous variables is based on calculations of quadratic entropies for each plot rather than discrete assignments of presence-absence of a functional group, or number of species.
| Factorial design | Complete design | TropiDEP design, discrete values | TropiDEP design, continuous values | |
|---|---|---|---|---|
| PDa Performance | 1 | 89% | 78% | 36% |
| FDcb Performance | 1 | 95% | 97% | 44% |
| FDnb Performance | 1 | 95% | 97% | 67% |
| Corr(PD, FDc) | 0 | –0.27 | –0.21 | –0.44 |
| Corr(PD, FDn) | 0 | –0.27 | –0.21 | –0.37 |
| Corr(FDc, FDn) | 0 | 0.10 | 0.03 | –0.01 |
| N (plot number) | 8 | 12 | 41 | 41 |
|
|
||||
| Advantages | Most efficient | Can verify linearity | Considers continuous factors | |
| Disadvantages | Does not verify linearity | Less efficient | ||