Ecological Archives E096064A8
Xubing Liu, Rampal S. Etienne, Minxia Liang, Yongfan Wang, and Shixiao Yu. 2015. Experimental evidence for an intraspecific JanzenConnell effect mediated by soil biota. Ecology 96:662–671. http://dx.doi.org/10.1890/140014.1
Appendix H. (Multivariate) odds ratio regression between survival log odds ratio and nine characteristic dissimilarities.
Table H1. Results of bestfit multivariate log odds ratio regressions with log odds ratio (field soil treated pots vs. sterile soil pots) as the response variable and the characteristic dissimilarities or the principal components as explanatory variables.
Variable 
AICc 
R² 
Estimate 
1 SE of estimate 
t 
P 
Castanopsis fissa 

1. Bestfir model with the variable dissimilarities 
33.10 
0.230 







1.144 
0.152 
7.5 
<0.001 



0.860 
0.328 
2.6 
0.015 
2. Bestfir model with the principal components 
32.11 
0.317 







0.843 
0.098 
8.6 
<0.001 



0.102 
0.042 
2.4 
0.025 



0.189 
0.135 
1.4 
0.177 



1.420 
1.003 
1.4 
0.172 
Canarium album 

3. Bestfir model with the variable dissimilarities 
35.70 
0.450 







1.327 
0.182 
7.3 
<0.001 



0.551 
0.346 
1.6 
0.127 



0.619 
0.350 
1.8 
0.093 



1.040 
0.441 
2.4 
0.029 



0.852 
0.588 
1.4 
0.163 
4. Bestfir model with the principal components 
37.71 
0.451 




(Intercept) 


0.790 
0.087 
9.0 
<0.001 



0.154 
0.044 
3.5 
0.002 



0.293 
0.179 
1.6 
0.117 



0.499 
0.249 
1.4 
0.167 
Notes: Log odds ratios were calculated by comparing survival in field soil treated pots vs. sterile soil pots (n = 30 seedlings per treatment). Multivariate log odds ratio regression models were fitted against all the nine characteristic dissimilarities or all the nine principal components, the bestfir model with a lowest AICc value was selected by a backward stepwise algorithm.
Table H2. Results of bestfit log odds ratio regressions with log odds ratio (soil filtrate treated pots vs. sterile soil pots) as the response variable and the characteristic dissimilarities or the principal components as explanatory variables.
Variable 
AICc 
R² 
Estimate 
1 SE of estimate 
t 
P 
Castanopsis fissa 

1. Bestfir model with the variable dissimilarities 
33.57 
0.478 







1.092 
0.200 
5.5 
<0.001 



2.260 
1.098 
2.1 
0.053 



1.403 
0.547 
2.6 
0.018 



3.268 
1.066 
3.1 
0.006 



0.985 
0.389 
2.5 
0.020 
2. Bestfir model with the principal components 
31.66 
0.437 




(Intercept) 


0.875 
0.097 
9.0 
<0.001 



0.115 
0.042 
2.7 
0.013 



0.094 
0.071 
1.3 
0.201 



0.225 
0.134 
1.7 
0.108 



1.683 
0.908 
1.9 
0.079 
Canarium album 

3. Bestfir model with the variable dissimilarities 
21.58 
0.442 







1.148 
0.253 
4.5 
<0.001 



3.082 
1.774 
1.7 
0.100 



3.254 
1.693 
1.9 
0.072 



1.134 
0.484 
2.3 
0.031 



0.965 
0.607 
1.6 
0.130 



2.668 
1.174 
2.3 
0.036 



2.490 
1.140 
2.2 
0.043 



0.856 
0.465 
1.8 
0.083 
4. Bestfir model with the principal components 
25.94 
0.405 







0.811 
0.109 
7.4 
<0.001 



0.131 
0.054 
2.4 
0.026 



0.257 
0.164 
1.6 
0.133 



0.356 
0.223 
1.6 
0.126 



0.725 
0.434 
1.7 
0.111 
Notes: Log odds ratios were calculated by comparing survival in soil filtrate treated pots vs. sterile soil pots (n = 30 seedlings per treatment). Multivariate log odds ratio regression models were fitted against all the nine characteristic dissimilarities or all the nine principal components, the bestfir model with a lowest AICc value was selected by a backward stepwise algorithm.
Fig. H1. Odds ratio regression between survival log odds ratio and trait dissimilarities for Castanopsis fissa seedlings. Log odds ratios were calculated by comparing survival in field soil treated pots vs. sterile soil pots (first and third columns) or soil filtrate treated pots vs. sterile soil pots (second and fourth columns) (n = 30 seedlings per treatment). For each trait/variable, the dissimilarity between any two populations was defined as their standardized absolute difference d_{iAB} (see Data analysis section for details). Logtransformed survival odds ratios above 0 indicate positive effects and log odds ratios that are significantly lower than 0 indicate negative effects. Bestfit odds ratio regression lines are shown, P values indicate significant departure from no relationship with trait dissimilarities. DBH: diameter at breast height; SLA: specific leaf area, LNC: leaf nitrogen concentration.
Fig. H2. Odds ratio regression between survival log odds ratio and trait dissimilarities for Canarium album seedlings. Log odds ratios were calculated by comparing survival infield soil treated pots vs. sterile soil pots (first and third columns) or soil filtrate treated pots vs. sterile soil pots (second and fourth columns) (n = 30 seedlings per treatment). For each trait/variable, the dissimilarity between any two populations was defined as their standardized absolute difference d_{iAB} (see Data analysis section for details). Logtransformed survival odds ratios above 0 indicate positive effects and log odds ratios that are significantly lower than 0 indicate negative effects. Bestfit odds ratio regression lines are shown, P values indicate significant departure from no relationship with trait dissimilarities. DBH: diameter at breast height; SLA: specific leaf area, LNC: leaf nitrogen concentration.
Fig. H3. Odds ratio regression between survival log odds ratio and trait dissimilarities measured on seedlings.