Changepoint alternatives to the NOAEL

The onset-of-trend changepoint model is applied to the dose-response setting encountered in regulatory toxicology experiments with continuous responses. An investigation of the least squares criterion over each of the dose intervals provides insights into the distribution of the changepoint estimate for small dose group sample sizes and a simple computational procedure for finding this estimate. A new accessible proof of the consistency of this estimate as dose group size increases is developed. For normally distributed data, an asymptotic procedure is developed for constructing a lower confidence limit on the changepoint. The performance of the changepoint estimate and the lower confidence limit are compared directly to the no-observed-adverse-effect-leveI (NOAEL) via a large simulation study and an example dataset. The lower confidence limit on the changepoint is shown to be a better regulatory quantity than the NOAEL from a statistical perspective.