Predator‐proof boma disrepair 199
nearest predator-proof boma and boma circumference). The variance inflation factor values for the remaining three predictors not included in the initial model (clustering of traditional bomas, distance to nearest protected area and months since construction)were 1.359–2.607, showingmod- erate or weak levels of correlation (coefficient range 0.337– 0.308). Weadded these predictors one at a time to the initial model. In addition, we fitted a null model (intercept-only). We conducted four separate analyses using, as the de- pendent variable, the proportion of damaged components (Supplementary Table 1, Analysis 1a), the proportion of damaged gates (Supplementary Table 1, Analysis 1b), the proportion of damaged posts (Supplementary Table 1, Analysis 1c) and the proportion of damaged chain-link fences (Supplementary Table 1, Analysis 1d). We employed a similar procedure for the second dataset
(n = 47), including four non-correlated predictors with the lowest variance inflation factor values (,1.2) in the initial model (homestead size, distance to nearest predator-proof boma, clustering of traditional bomas and livestock density). Predictors not included in the initialmodel (distance to pro- tected area, extent of conflict and months since construc- tion) also had low variance inflation factor values (1.1–1.3) but were weakly correlated with some variables (coefficient range 0.350–0.176). Based on the criteria defined above, we fitted five models
for each analysis. To assess whether the number of variables in each model affected model quality, we compared the Akaike information criterion (AIC) of the null model to the AIC of each model; in addition, we examined binned scatter plots of the relationships between the predicted values and the observed values. This procedure helped us to select final models.
Model evaluation
Amongst all models fitted using the main dataset (n = 86)and proportions of damaged components as the dependent vari- able (Supplementary Table 1,Analysis 1a), the fit of the initial model (five predictors) was the best, as shown by the AIC va- lues (Supplementary Table 1). The inclusion of an additional predictor, clustering of traditional bomas, also increased the fit, but the fit of models with distance to nearest protected area and months since construction decreased relative to the other models. However, the binned scatter plots of the predicted values against the observed values were similar in all models in always indicating a good fit of the data, thus suggesting that the inclusion of additional variables did not have a major influence on the quality of the models. We therefore included all variables in the final model, given the good fit (Supplementary Fig. 1a). Regardless of the number of predictors included in the different models, the effects of the same two variables
(livestock density and boma post type) were consistently sig- nificant and the sign of their regression coefficients did not
change (−0.0368 to−0.0345 and−3.4183 to−2.9413); other predictors always failed to reach statistical significance. Similar patterns emerged when using proportion of damaged posts as the dependent variable (Supplementary Table 1, Analysis 1c); the final model with all variables also fitted the data well (Supplementary Fig. 1c). The fit of all models built using proportion of damaged chain-link fences as the dependent variable decreased compared to that of the null model (Supplementary Table 1, Analysis 1d); nevertheless, all variables were retained in the final model, which provided a reasonable fit to the data (Supplementary Fig. 1d). Proportion of damaged gates was also used as a dependent variable, but this dataset was poor (most values for the dependent variable were 0) and we obtained no well-fitting model. The initial model built with five variables poorly fit- ted the data (Supplementary Fig. 1b) and additional vari- ables did not improve the model (Supplementary Table 1, Analysis 1b). For the second dataset (n = 47), the binned scatter plot
for the initial model (four predictors) constructed using proportion of damaged components as the dependent vari- able suggested a suitable fit of the data (Supplementary Fig. 2a). The AIC values for models with additional variables indicated a decrease of fit relative to the initial model (Supplementary Table 1, Analysis 2a). However, the same predictor was consistently significant in all models, always having a positive influence (regression coefficient range 0.000164–0.000171), whereas all other predictors remained statistically non-significant. Given the limited sample size, we did not include any additional predictors in the final model. Patterns were similar to those of Analysis 2c, which included proportion of damaged posts as the depen- dent variable (Supplementary Table 1 & Supplementary Fig. 2c), but predictors were statistically non-significant in all models except for one case where the model (with six pre- dictors) yielded P = 0.0496 for distance to nearest predator- proof boma. However, given the limited sample size, we did not consider this as the final model, to avoid overfitting. For models including proportion of damaged gates and propor- tion of damaged chain-link fences as dependent variables, binned scatter plots of the initial models indicated a con- siderable lack of fit (Supplementary Fig. 2b, d); other pre- dictors did not change this pattern (Supplementary Table 1, Analyses 2b,d).
Results
Boma characteristics The mean numbers of livestock per boma were 383.4 ± SD 333.0 (n = 88) at the time of construction and 226.3 ± SD
Oryx, 2023, 57(2), 196–204 © Born Free Foundation and the Author(s), 2022. Published by Cambridge University Press on behalf of Fauna & Flora International doi:10.1017/ S0030605321001642
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72 |
Page 73 |
Page 74 |
Page 75 |
Page 76 |
Page 77 |
Page 78 |
Page 79 |
Page 80 |
Page 81 |
Page 82 |
Page 83 |
Page 84 |
Page 85 |
Page 86 |
Page 87 |
Page 88 |
Page 89 |
Page 90 |
Page 91 |
Page 92 |
Page 93 |
Page 94 |
Page 95 |
Page 96 |
Page 97 |
Page 98 |
Page 99 |
Page 100 |
Page 101 |
Page 102 |
Page 103 |
Page 104 |
Page 105 |
Page 106 |
Page 107 |
Page 108 |
Page 109 |
Page 110 |
Page 111 |
Page 112 |
Page 113 |
Page 114 |
Page 115 |
Page 116 |
Page 117 |
Page 118 |
Page 119 |
Page 120 |
Page 121 |
Page 122 |
Page 123 |
Page 124 |
Page 125 |
Page 126 |
Page 127 |
Page 128 |
Page 129 |
Page 130 |
Page 131 |
Page 132 |
Page 133 |
Page 134 |
Page 135 |
Page 136 |
Page 137 |
Page 138 |
Page 139 |
Page 140
