Skip to main content
. 2010 Feb 19;163(3):805–813. doi: 10.1007/s00442-010-1569-5

Table 1.

Basis sets for the partial independence constraints implied by each of the five alternative models.

Model A Model B Model C Model D
Basis set r (P) Basis set r (P) Basis set r (P) Basis set r (P)
(S,T)|{Ø} −0.03 (0.721) (S,T)|{Ø} −0.03 (0.721) (S,T)|{Ø} −0.03 (0.721) (S,T)|{Ø} −0.03 (0.721)
(S,R)|{T,P} −0.02 (0.851) (S,R)|{T,N} 0.17 (0.090) (S,R)|{T,P,N} −0.02 (0.869) (S,R)|{T,P} −0.02 (0.851)
(T,P)|{S} −0.03 (0.792) (T,P)|{S} −0.03 (0.792) (T,P)|{S} −0.03 (0.792) (T,P)|{S,N} −0.07 (0.468)
(T,N)|{S} −0.11 (0.301) (T,N)|{S} −0.11 (0.301) (T,N)|{S} −0.11 (0.301) (T,N)|{S} −0.11 (0.301)
(P,N)|{S} −0.39 (<0.001) (P,N)|{S} −0.39 (<0.001) (P,N)|{S} −0.39 (<0.001) (N,R)|{S,T,P} 0.01 (0.947)
(N,R)|{S,T,P} 0.01 (0.947) (P,R)|{S,T,N} 0.42 (<0.001)
12 df, C = 23.69 (0.022) 12 df, C = 50.29 (<0.001) 10 df, C = 23.54 (0.009) 10 df, C = 5.01 (0.890)
Model E Model F Model G Model H
Basis set r (P) Basis set r (P) Basis set r (P) Basis set r (P)
(S,T)|{Ø} −0.03 (0.721) (S,T)|{Ø} −0.03 (0.721) (S,T)|{Ø} −0.03 (0.721) (S,T)|{Ø} −0.03 (0.721)
(S,R)|{T,N} 0.17 (0.090) (S,R)|{T,P,N} −0.02 (0.869) (T,P)|{S,N} −0.07 (0.468) (T,P)|{S} −0.03 (0.792)
(T,P)|{S,N} −0.07 (0.468) (T,P)|{S,N} −0.07 (0.468) (T,N)|{S} −0.11 (0.301) (T,N)|{S} −0.11 (0.301)
(T,N)|{S} −0.11 (0.301) (T,N)|{S} −0.11 (0.301) (P,R)|{S,T,N} 0.42 (<0.001) (P,N)|{S} −0.39 (<0.001)
(P,R)|{S,T,N} 0.42 (<0.001) (N,R)|{S,T} −0.17 (0.092) (P,R)|{S,T} 0.45 (<0.001)
(N,R)|{S,T} −0.17 (0.092)
10 df, C = 31.62 (<0.001) 8 df, C = 4.86 (0.772) 10 df, C = 31.59 (<0.001) 12 df, C = 53.64 (<0.001)

The hypothesized relationship among variables can be seen in Fig. 1. S Species richness, T soil temperature, P biomass, N average community leaf N concentration, R soil respiration

The notation (X,Y)|{A,B,…} means that variables x and y are hypothesized to be probabilistically independent conditional on the set of variables {A,B,…}. Ø represents the null (empty) set. Pearson’s partial correlation coefficient (r), and P (assuming the null hypothesis) in parentheses, are given for each conditional independence claim; the overall model is tested with Fisher’s C statistic

P < 0.05 indicated in bold, rejecting the hypothesized independence relationship