Appendix D. Importance of habitat variables in model results.
TABLE D1. Importance of habitat variables in model results. Importance was assessed differently for different model types. LR: number of replicates in which the variable was retained after stepwise removal; values in parentheses are mean regression coefficients (and sign) among replicates. GARP: For each replicate model, we created 50 GARP models with each of the 12 habitat variables removed (600 models/replicate). We then used t tests to compare the accuracy (Sensitivity) of models that included the variable of interest (550 models/replicate) with those in which the variable was removed (50 models/replicate). Values are mean t statistics (and SEs) for the five replicates of each species. The magnitude of a positive mean t statistic reflected the degree to which a variable improved model accuracy. ENFA: mean loading on the first factor, indicating the difference in a variable for points representing species' occurrences and the median of the study area.

— Geranium — 
— Hedera — 
— Ilex — 

Habitat variable 
LR 
GARP 
ENFA 
LR 
GARP 
ENFA 
LR 
GARP 
ENFA 
Intercept 
(25.56) 
 
 
(66.62) 
 
 
(90.58) 
 
 
Annual precipitation 
5 
0.03 
0.55 

0.44 
0.57 
4 
0.40 
0.55 
Conifer cover 
5 
0.05 
0.25 

0.85 
0.07 

0.33 
0.09 
Distance to water 
5 
0.32 
0.30 

0.13 
0.08 

0.82 
0.09 
Frost days 
5 
0.62 
0.30 
5 
1.86 
0.45 
5 
1.59 
0.41 
Heat load 

0.80 
0.10 

0.73 
0.01 

0.67 
0.08 
Potential radiation 
1 
0.67 
0.08 
1 
0.51 
0.01 

1.69 
0.07 
Precipitation frequency 
5 
2.14 
0.48 

0.05 
0.24 
5 
2.71 
0.30 
Radiation 
5 
0.86 
0.01 
5 
1.53 
0.13 
5 
0.94 
0.11 
Slope 

1.18 
0.31 

0.13 
0.40 

3.77 
0.45 
Vegetation cover 

0.79 
0.08 

0.01 
0.22 

0.46 
0.07 
Humidity 
5 
1.43 
0.14 

1.37 
0.32 
5 
0.65 
0.28 
Topographic moisture index 

0.59 
0.30 

0.83 
0.22 

0.39 
0.27 
Marginality 
 
 
1.17 
 
 
1.02 
 
 
1.02 
Specialization 
 
 
2.07 
 
 
6.24 
 
 
4.39 
Residual deviance 
189.66 
 
 
15.09 
 
 
27.21 
 
 