Table 5. Logistic regression models, as in Table 4, explaining the presence or absence of Capercaillie signs at individual sample plots (n = 2340) from HSI scores at various scales. In contrast to the models in Table 4, the landscape variables HSI1NN (extent of excellent HSI clusters) and FOREST5KM (percentage of forest cover within mountain ranges) were entered instead of the categorical variable AREA (study area) (Models 1–4, 6). These landscape-scale variables offered the best explanation for Capercaillie abundance at the scale of study areas (Model 5, variables entered: AREA-HSI year, CV-AREA-HSI year, HSI1NN, HSI11, FOREST5KM, FOREST10KM). All models resulted from backward elimination of variables.

 
Model (scale)
and variables
B
1 SE
df
Odds ratio exp(B)
P
Change in deviance
Residual deviance
(-2log-likelihood)
 
 
1) Sample plot
 
700.7

    HSIsu
4.40
0.45
1
81.6
0.000
102.2
  
    HSI1NN
0.97
0.26
1
2.63
0.000
16.5
  
    FOREST5KM
0.02
0.01
1
1.02
0.036
4.6
  
    Constant
- 8.02  
0.98
  
0.000
    
 
 
2) Sample plot and 8 nearest neighbors
 
731.4

    NN-HSIyear
6.35
0.76
1
573.2
0.000
77.9
  
    HSI1NN
1.05
0.27
1
2.86
0.000
16.7
  
    Constant
- 6.88  
0.49
  
0.001
    
 
 
3) 100-ha quadrat
 
731.4

    100-HSIyear
6.67
0.80
1
784.4
0.000
72.7
  
    HSI1NN
0.86
0.27
1
2.37
0.002
10.6
  
    Constant
- 6.70  
0.45
  
0.000
    
 
 
4) 400-ha quadrat
 
765.1

    400-HSIyear
4.39
1.46
1
80.4
0.003
9.15
  
    CV 400-HSIyear
- 0.02  
0.01
1
0.98
0.012
6.86
  
    HSI1NN
0.97
0.26
1
2.63
0.000
14.5
  
    Constant
- 4.78  
0.90
  
0.000
    
 
 
5) 2000-ha quadrat
 
803.0

    HSI1NN
0.98
0.21
1
2.678
0.000
26.2
  
    FOREST5KM
0.02
0.01
1
1.023
0.022
5.5
  
    Constant
- 6.04  
0.87
  
0.002
    
 
 
6) All scales
 
691.2

    HSIsu
3.36
0.52
1
28.90
0.000
40.23
  
    NN-HSIyear
3.23
0.88
1
25.34
0.000
14.20
  
    HSI1NN
1.03
0.28
1
2.80
0.000
14.74
  
    Constant
- 7.04  
0.49
  
0.001