Parameter estimates and 95% profile likelihood-based confidence intervals (95% CI) (or 95% Bayesian credible intervals (BCI)) for analysis of curvature in fits of excess odds ratio model (1) with or without adjustment for dose errors using regression calibration, for various sets of doses.

<p>^aunless otherwise stated all <i>p</i>-values refer to the improvement in fit of the current row in the Table with that of the model fitted in the row immediately above.</p><p>^b<i>p</i>-value of improvement in fit compared with a model with linear-exponential dose terms.</p><p>^c<i>p</i>-value of improvement in fit compared with a model with no dose terms.</p><p>^ddeviance information criterion [<a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0139826#pone.0139826.ref024" target="_blank">24</a>].</p><p>^eposterior distribution maximum probability estimate.</p><p>^f95% Bayesian credibility interval (BCI).</p><p>All models have underlying rates adjusted for age (treated categorically), gender and oblast.