Events

Date: 2 April 2026   Time: 14:00 - 15:00     

Location: Hybrid: MB503, SMS, QMUL, or via the Teams link below

We consider testing the goodness of fit of semiparametric regression models, such as generalised additive models, partially linear models, or quantile additive regression models. We propose an approach that involves first splitting the data in two parts. On one part, we "hunt" for any signal that may be present in the score-type residuals following a fit of the model. On the remaining data, we test for the presence of the potential signal thus found. For the first hunting stage of the procedure, our framework allows for carrying this out based on a user-chosen flexible regression method, such as a random forest. The method is thus able to leverage the power of modern machine learning methods to detect complex alternatives. A challenge with the testing step is that any first-order bias in the residuals may lead to rejection under the null. We therefore employ a debiasing step which we show amounts to performing a particular weighted least squares regression. We show that the type I error may be controlled under relatively mild conditions, and that we have power under alternatives where with high probability the hunted signal is correlated with the true signal present in the score residuals.