Spatiotemporal dynamics of Puumala hantavirus in suburban reservoir rodent populations
Alexandre Dobly, Chloé Yzoard, Christel Cochez, Geneviève Ducoffre, Marc Aerts, Stefan Roels and Paul Heyman
Journal of Vector Ecology (2012), 37, 276-283
Additional Tables
Table S1. Determination of factors that influence seroprevalence in bank voles by univariate logistic regression: estimate coefficient, correcteda estimate standard error (sec), correcteda p-value, estimate odds ratio (OR), correcteda confidence interval (CIc), Akaike’s Information Criterion (AIC)
Variable / Estimate / sec / p-valuec / OR / CIc / AICSite 2 / 10.134 / 0.5329 / 0.0572 / 2.755 / (0.969,7.829) / 500.336
Site 3 / 0.6123 / 0.5198 / 0.2388 / 1.845 / (0.666,5.110)
Year / 0.7328 / 0.4733 / 0.1216 / 2.081 / (0.823,5.261) / 502.852
Seasonb / 16.985 / 0.4086 / <.0001*** / 5.466 / (2.454,12.176) / 466.114
Density / 0.0254 / 0.0057 / <.0001*** / 1.026 / (1.014,1.037) / 465.980
Sexc / -0.1155 / 0.4681 / 0.8050 / 0.891 / (0.356,2.230) / -
Weightd / 0.0659 / 0.0268 / 0.0141* / 1.068 / -1.013,13 / -
Max. dist.e / 0.0143 / 0.0076 / 0.0583 / 1.014 / (0.999,1.030)
Max. areaf / 0.0009 / 0.0010 / 0.3469 / 1.001 / (0.999,1.003)
Index 1g / 0.0117 / 0.0039 / 0.0028** / 1.012 / (1.004,1.020)
Index 2g / 0.0005 / 0.0003 / 0.0821 / 1.000 / (1.000,1.001)
a Corrected estimate standard error, p-value, and 95% confidence interval: all univariate logistic regression models show lack of fit using the deviance goodness of fit statistic (not shown), leading to overdispersion. As a consequence standard errors are underestimated, p-values are too low and confidence intervals are too narrow. The corrected versions are corrected for overdispersion.
b Season was coded as 1 for spring and 0 for autumn.
c Sex was coded as 1 for males and 0 for females. For 13 voles sex was unknown and these observations were deleted. This implies that the AIC value of the model with sex cannot be compared with the other AIC values, and therefore it is not shown.
d As weight was missing for two observations, the AIC value of the model with weight cannot be compared with the other AIC values, therefore it is not shown. Moreover, since weight was included as a continuous variable the deviance statistic can no longer be used for checking lack of fit and for correcting estimates for standard errors, p-values or confidence intervals.
e Maximal distance between two captures: Known only for 242 voles captured at least in two traps.
f Maximal area between at least three captures: Known only for 115 voles captured at least in three traps.
g Indexes are available only for 85 voles captured at least in three traps and with known weight; Index 1 is in m/g (of body weight); Index 2 is in m2/g (of body weight).
Table S2. Determination of the manner significant factors influence seroprevalence in bank voles by multiple covariate logistic regressiona. Upper panel: final model based on all data. Lower panel: final model based on subset with index values.
Intercept / -18.281 / 0.3591 / <.0001***
Site2 / -18.854 / 10.739 / 0.0792
Site3 / -0.2356 / 0.5195 / 0.6502
Season / 0.4538 / 0.4343 / 0.2961
Season × Site2 / 35.244 / 11.254 / 0.0017***
Season × Site3 / 13.586 / 0.6167 / 0.0276*
Intercept / -44.280 / 10.301 / <.0001***
Index1 / 0.0129 / 0.0044 / 0.0032***
Season / 15.242 / 0.7305 / 0.0369*
a OR’s are no longer reported as interaction effects (of season and site) complicate interpretation. estimate = estimate coefficient, se = standard error
Table S3. Determination of the effect of season and site on vole weight by multiple regression for log(weight): estimate coefficient, estimate standard error (se), p-value
Intercept / 29.596 / 0.0326 / <.0001***
Site2 / -0.0382 / 0.0337 / 0.2587
Site3 / -0.0731 / 0.0365 / 0.0459*
Season / 0.2277 / 0.0399 / <.0001***
Year / 0.0542 / 0.0325 / 0.0957
Season × Site2 / 0.0859 / 0.0428 / 0.0456*
Season × Site3 / 0.0489 / 0.0456 / 0.2846
Season × Year / -0.1965 / 0.0426 / <.0001***
Log of weight was used to get normally distributed residuals.
3