Statistical inference and visualization in scale-space using local likelihood

SiZer (SIgnificant ZERo crossing of the derivatives) is a graphical scale-space visualization tool that allows for exploratory data analysis with statistical inference. Various SiZer tools have been developed in the last decade, but most of them are not appropriate when the response variable takes discrete values. In this paper, we develop a SiZer for finding significant features using a local likelihood approach with local polynomial estimators. This tool improves the existing one (Li and Marron, 2005) by proposing a theoretically justified quantile in a confidence interval using advanced distribution theory. In addition, we investigate the asymptotic properties of the proposed tool. We conduct a numerical study to demonstrate the sample performance of SiZer using Bernoulli and Poisson models using simulated and real examples.