Sickness behaviour reduces network centrality in wild vampire bats

Simon P Ripperger; Sebastian Stockmaier; Gerald G Carter

Defining network edges requires deciding what minimum proximity or duration 130 constitutes an 'association'. With proximity sensor data, this definition is flexible but requires a 131 trade-off between maximizing the sample size of observations and filtering for observations that 132 are more meaningful (e.g. closer proximities or longer durations). We inspected how the size 133 and precision of the LPS effect changed with variations in how networks were constructed. To 134 do this, we resampled our data using different definitions of association, then we plotted 135 changes in the number of observations and the treatment effect size (defined as the 136 unstandardized model coefficient during the treatment period). As a measure of relative 137 detectability, we used the p-value from the parametric linear model. To investigate the effect of 138 minimum encounter duration, we defined association at one proximity index (85% as in our 139 original analysis), but filtered encounters using several values of minimum duration that varied 140 . CC-BY-NC-ND 4.0 International license author/funder. It is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.03.30.015545 doi: bioRxiv preprint from 0 -1200 s. To investigate the effect of minimum encounter proximity, we set no minimum 141 duration (as in our original analysis), but filtered encounters using a proximity index threshold 142 that varied across several values from 76 -98 %. Proximity index thresholds below 76% are not 143 informative given the roost size; thresholds over 98% use less than 2% of the data. Finally, to 144 assess if the LPS treatment effect size was robust across different proximity thresholds, when 145 controlling for number of observations, we used several proximity index thresholds varying from 146 75% to 94% but we randomly sub-sampled the same number of observations in each case 147 The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi. org/10.1101 org/10. /2020 Results 153 154