Title: Testing goodness-of-fit with complex data
Authors: Ming-Yen Cheng - Hong Kong Baptist University (Hong Kong) [presenting]
Abstract: In the emerging era of big data, it occurs often that both the dimensionality and complexity of the data are high. Although parametric models enjoy good efficiency and interpretability if they are correctly specified, they suffer from large bias if they are incorrect. Nonparametric models are flexible in allowing the data to speak for the underlying structure and can serve as useful tools for model checking. However, they run into the curse of dimensionality problem and the variance dominates. In addition, in some cases identification is a problem. Semiparametric models offer a trade-off between parametric and nonparametric models and enjoy advantages of both of the two. Therefore they are particularly useful in analysis of complex data. We will focus on goodness-of-fit testing for some semiparametric regression models. Theoretical and numerical results, along with applications to data sets coming from medical and climate studies will be given to demonstrate the efficacy and advantages of the proposed methods.