Inference for local parameters in convexity constrained models

We consider the problem of inference for local parameters of a convex regression function based on observations from a standard nonparametric regression model, using the convex least squares estimator (LSE) . For , the local parameters include the pointwise function value , the pointwise derivative , and the anti-mode (i.e., the smallest minimizer) of . The existing limiting distribution of the estimation error depends on the unknown second derivative , and is therefore not directly applicable for inference. To circumvent this impasse, we show that the following locally normalized errors (LNEs) enjoy pivotal limiting behavior: Let be the maximal interval containing where is linear. Then, under standard conditions, \binom{ \sqrt{n(\widehat{v}(x_0)-\widehat{u}(x_0))}(\widehat{f}_n(x_0)-f_0(x_0)) }{ \sqrt{n(\widehat{v}(x_0)-\widehat{u}(x_0))^3}(\widehat{f}_n'(x_0)-f_0'(x_0))} \rightsquigarrow \sigma \cdot \binom{\mathbb{L}^{(0)}_2}{\mathbb{L}^{(1)}_2}, where is the sample size, is the standard deviation of the errors, and are universal random variables. This asymptotically pivotal LNE theory instantly yields a simple tuning-free procedure for constructing CIs with asymptotically exact coverage and optimal length for and . We also construct an asymptotically pivotal LNE for the anti-mode of , and its limiting distribution does not even depend on . These asymptotically pivotal LNE theories are further extended to other convexity/concavity constrained models (e.g., log-concave density estimation) for which a limit distribution theory is available for problem-specific estimators.
View on arXiv