Thesis (M.S., Statistical Sciences)--University of Idaho, June 2014 | Logistic regression is often used to predict the probability of an event, but it assumes perfect test sensitivity and specificity. However, most tests are not perfectly sensitive and specific. The prediction interval is an estimate of the interval in which the future observations will fall, and it can be applied to study the impact of imperfect test.
Logistic regression is applied to the study of invasive species, such as New Zealand mudsnails, which will potentially harm biodiversity and affect biotic homogenization.
In this study, we investigate how risk prediction in logistic regression on New Zealand mudsnail is affected by imperfect tests, using a Bayesian approach. The results show that the changes of mean sensitivity clearly affect the prediction interval width at a low temperature, but the effects of changes of mean specificity and the weights of prior distributions of sensitivity and specificity were less clear at both temperatures examined.