Impact of fitting age in general linear models testing the strength of association between genotype and reactor status for the SICCT test of exposure to bovine tuberculosis.

<p>Each model is fitted with response variable reactor status (0 = positive test, 1 = negative test) and group (either breed, ‘B’, or STRUCTURE group, ‘S’, fitted as a factor) and genotype (‘G’, fitted as a factor) either with or without age at slaughter (‘A’, fitted as a covariate. Models either included (*) or did not include (+) all possible second order interactions between terms. Significance was assessed by repeatedly resampling genotype within breed/group without replacement and refitting the model. As in <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0018806#pone-0018806-t002" target="_blank">Table 2</a>, significant P-values (P<0.05, uncorrected for multiple tests) are highlighted in bold. P-values of 0 indicate none of 100,000 randomizations exceeded the initial observed value.