Efficient Estimation of Pathwise Differentiable Target Parameters with the Undersmoothed Highly Adaptive Lasso (original) (raw)

View PDF

Abstract:We consider estimation of a functional parameter of a realistically modeled data distribution based on observing independent and identically distributed observations. We define an mmm-th order Spline Highly Adaptive Lasso Minimum Loss Estimator (Spline HAL-MLE) of a functional parameter that is defined by minimizing the empirical risk function over an mmm-th order smoothness class of functions. We show that this mmm-th order smoothness class consists of all functions that can be represented as an infinitesimal linear combination of tensor products of leqm\leq mleqm-th order spline-basis functions, and involves assuming mmm-derivatives in each coordinate. By selecting mmm with cross-validation we obtain a Spline-HAL-MLE that is able to adapt to the underlying unknown smoothness of the true function, while guaranteeing a rate of convergence faster than nāˆ’1/4n^{-1/4}nāˆ’1/4, as long as the true function is cadlag (right-continuous with left-hand limits) and has finite sectional variation norm. The m=0m=0m=0-smoothness class consists of all cadlag functions with finite sectional variation norm and corresponds with the original HAL-MLE defined in van der Laan (2015).
In this article we establish that this Spline-HAL-MLE yields an asymptotically efficient estimator of any smooth feature of the functional parameter under an easily verifiable global undersmoothing condition. A sufficient condition for the latter condition is that the minimum of the empirical mean of the selected basis functions is smaller than a constant times nāˆ’1/2n^{-1/2}nāˆ’1/2, which is not parameter specific and enforces the selection of the L_1L_1L_1-norm in the lasso to be large enough to include sparsely supported basis. We demonstrate our general result for the m=0m=0m=0-HAL-MLE of the average treatment effect and of the integral of the square of the data density. We also present simulations for these two examples confirming the theory.

Submission history

From: Weixin Cai [view email]
[v1] Wed, 14 Aug 2019 16:28:37 UTC (2,759 KB)
[v2] Fri, 2 Jul 2021 17:32:36 UTC (5,521 KB)