Table 4. Prediction of difference in swelling size between initial and final measurements at the bovine tuberculin injection site (db) by ‘22’ genotype.

Count model (Poisson)
Age	1.0 (1.0–1.0)	−0.254	0.7998
b1	1.0 (0.86–1.3)	0.560	0.5758
p22	1.6 (0.82–4.0)	1.985	0.0472
Zero-inflation model (Binomial)
	Odds Ratio	z value	Pr(>|z|)
Age	1.0 (1.0–1.0)	−2.191	0.0284
b1	0.91 (0.76–1.1)	−1.280	0.2005
p22	1.44 (0.57–4.05)	0.841	0.4005

Incident risk and odds ratios for both components of a zero-inflated Poisson model fitted to the bovine difference (db ∼ Age + b1 + p22). Odds and incident risk ratios (from the Poisson count model and binomial zero inflation terms respectively) are presented to two significant figures, along with 95% confidence intervals calculated from 10000 parametric bootstraps. Significant effects at the 95% level are highlighted in bold. While the age co-efficient is highly significant with the Poisson portion of the model, the p22 effect is only marginally significant for da. The marginal significance of the p22 effect is further emphasised by the variability in the bootstrapped confidence interval, which constitutes a more conservative test.