Interpreting data in the face of competing explanations: assessing the hypothesis that observed spontaneous clearance of Helicobacter pylori was all measurement error

Methods We address the logic of the implicit argument that the transient infections were an artifact and we demonstrate a simple likelihood calculation to assess the plausibility of competing explanations. We calculate the likelihood of observing our data based on a range of clearance and measurement error rates and then how this updates a set of prior beliefs. Results The likelihood calculations and resulting posterior probabilities show strong support for the hypothesis of spontaneous clearance, after allowing for measurement error, even starting with a very high prior probability of no spontaneous clearance. The scenario Perry and Parsonnet present is incompatible with our data, and thus not a plausible explanation for our observations. Attributing most observed transient infections to measurement error requires assuming a high false positive rate and a very low infection rate and/or a high false negative rate, alternatives that are not supported by evidence. Conclusions Acknowledgment of plausible levels of measurement error does not change the strong support our data provides for the hypothesis of frequent transient infection. Debate about competing explanations for observations should be accompanied by quantitative analysis that shows which is more plausible. We demonstrate one method for doing such analysis.