Variable

Estimate

Std. error

Exp(est)

Pvalue

Δ AIC


Intercept

2.30

0.84
 
0.006
 
Area shrubi 1 vs. 0*

1.05

0.27

2.86

<0.001

14

Area shrubi 2 vs. 0*

0.85

0.23

2.35

<0.001

Area shrubi 3 vs. 0*

0.79

0.23

2.21

<0.001

Meanarea_ p

0.45

0.12

1.40

<0.001

9

Meanarea _p^{2}

−0.11

0.03

<0.001

RHmean_{
OctMar
}

1.21

0.13

3.34

<0.001

88

BlackFrdays

0.92

0.16

2.50

<0.001

55

SnoStartDays

1.17

0.20

3.23

<0.001

34

NuFarms

2.67

0.31

2.27

<0.001

94

NuFarms^{2}

−1.85

0.21

<0.001

Red deer

1.28

0.22

3.59

<0.001

29

Pasture

−1.36

0.30

0.25

<0.001

18

RRSum_{
May
}

−0.18

0.15

0.59

0.224

46

$\mathrm{RRSu}{\mathrm{m}}_{\mathrm{May}}^{2}$

−0.35

0.06

<0.001

RRSum_{
Mar
}

−0.40

0.14

0.67

0.004

6

TIncr + 5 < Days_{
Jun
}

−1.11

0.21

0.40

<0.001

27

$\mathrm{TIncr}+5<\mathrm{Day}{\mathrm{s}}_{\mathrm{Jun}}^{2}$

0.19

0.04

<0.001

TMeanSD_{
Apr
}

0.43

0.22

1.93

0.047

5

$\mathrm{TMeanS}{\mathrm{D}}_{\mathrm{Apr}}^{2}$

0.22

0.08

0.004

TDecr ÷ 5 < Days_{Jan − Dec}

0.22

0.10

1.25

0.035

2

 Δ AIC denotes the change in AIC level obtained if excluding the relevant variable from the selected model. The continuous variables are scaled (before taking polynomials) to mean zero and variance one. The factor by which the odds of positive outcome are increased for each oneunit change in the variables are represented by the computed exp (estimates). For the polynomials the odds ratio is calculated only for an increase of one standard deviation from mean. ICC (Intraclass correlation; ratio of the variance between subjects over the total variance) for municipality was 0.33 and ICC for timespan was 0.36. The “Area shrubi” variable was categorized (in 4 equal parts defined by quartiles) to capture the nonlinear relationship (at logitscale) with the outcome. The variables Area shrubi, BlackFrDays and RRSumMar represent the rough grazing level.