Figure 3

(a) Effect display⁴⁴ for the calculated total dose (log₁₀-transformed) predictor of total number of birds (TNB), in the fitted GLMM. After controlling for the possible confounding variables, the total number of birds decreased by 22.6% per unit of log₁₀ total dose. Note that the vertical axis is on a scale of the linear predictor with axis tick marks labelled on the scale of the response (TNB). The decrease is highly significant (z = −4.06, p < 0.0001; unstandardised partial slope (SE) = −0.256 (0.063)). (b) Effect display for the calculated total dose (log₁₀-transformed) predictor of Simpson’s index of diversity, in the fitted LMM. After controlling for the possible confounding variables, the probability that two individuals belonged to different species increases linearly by 4.5% per unit of log₁₀ total dose. The increase is significant (t = 2.3, p < 0.025; unstandardised partial slope (SE) = −0.045 (0.020)); Grey bands in the panels represent the 95% confidence intervals around the fits.