Variables Associated with Strikes
The variability in train strikes to relative abundance and nine train or railroad design
variables (Table 9.1) indicated a negative binomial distribution for all species
models. Initial models did not indicate that autocorrelation was present between rail
segments and the most parsimonious models are reported (Table 9.3).
Table 9.2 Variation between years and railroad segments for (a) elk, (b) deer, and (c) bears along
the Canadian Pacific Railroad within the Banff and Yoho National Parks, 1989–2009
Factor
df
Sum Sq.
Mean Sq.
F
p
(a) Elk
Year
20
311.31
15.57
2.4
<0.001
Railroad segment
26
760.81
29.26
4.6
<0.001
Residuals
234
1489.85
6.37
(b) Deer
Year
20
74.19
3.71
3.7
<0.001
Railroad segment
26
67.65
2.60
2.6
<0.001
Residuals
234
233.37
1.00
(c) Bear
Year
20
17.19
0.86
3.2
<0.001
Railroad segment
26
22.34
0.86
3.2
<0.001
Residuals
234
63.27
0.27
Table 9.3 Significant (p < 0.05) parameter estimates explaining train strike rates with elk, deer,
and bears
Elk
Deer
Bears
b
SE
p
b
SE
p
b
SE
p
Explanatory variables
Y-intercept −1.063 0.782 <0.001 −1.917 0.751 <0.001
0.825 0.153 <0.001
RA rail
corridor
0.024 0.007 <0.001
0.069 0.014 <0.001
ROW mean
0.009 0.003
0.005
SPEED max
0.080 0.020 <0.001
0.055 0.017
Categorical variables
BARRIERc
0.458 0.170
0.007
Dispersion
(h)
1.5
3.8
20.0
Deviance
explained
(%)
41.7
28.3
82.042
Model-fit statistics are provided from maximum likelihood fits
148
B.P. Dorsey et al.