arXiv subjects
Publications and source records attributed to Fadoua Balabdaoui.
Based on the convex least-squares estimator, we propose two different procedures for testing convexity of a probability mass function supported on N with an unknown finite support. The procedures are shown to be asymptotically calibrated.
In this article, we revisit the problem of estimating the unknown zero-symmetric distribution in a two-component location mixture model, considered in previous works, now under the assumption that the zero-symmetric distribution has a log-concave density. When consistent estimators for the shift locations and mixing probability are used, we show that the nonparametric log-concave Maximum Likelihood estimator (MLE) of both the mixed density and that of the unknown zero-symmetric component are consistent in the Hellinger distance. In case the estimators for the shift locations and mixing probability are $\sqrt n$-consistent, we establish that these MLE's converge to the truth at the rate $n^{-2/5}$ in the $L_1$ distance. To estimate the shift locations and mixing probability, we use the estimators proposed by \cite{hunteretal2007}. The unknown zero-symmetric density is efficiently computed using the \proglang{R} package \pkg{logcondens.mode}.
In this article, we derive the weak limiting distribution of the least squares estimator (LSE) of a convex probability mass function (pmf) with a finite support. We show that it can be defined via a certain convex projection of a Gaussian vector. Furthermore, samples of any given size from this limit distribution can be generated using an efficient Dykstra-like algorithm.
We propose a new testing procedure for detecting localized departures from monotonicity of a signal embedded in white noise. In fact, we perform simultaneously several tests that aim at detecting departures from concavity for the integrated signal over various intervals of different sizes and localizations. Each of these local tests relies on estimating the distance between the restriction of the integrated signal to some interval and its least concave majorant. Our test can be easily implemented and is proved to achieve the optimal uniform separation rate simultaneously for a wide range of Hölderian alternatives. Moreover, we show how this test can be extended to a Gaussian regression framework with unknown variance. A simulation study confirms the good performance of our procedure in practice.
We show that the density of $Z=\mathop {\operatorname {argmax}}\{W(t)-t^2\}$, sometimes known as Chernoff's density, is log-concave. We conjecture that Chernoff's density is strongly log-concave or "super-Gaussian", and provide evidence in support of the conjecture.
The k-monotone classes of densities defined on (0, \infty) have been known in the mathematical literature but were for the first time considered from a statistical point of view by Balabdaoui and Wellner (2007, 2010). In these works, the authors generalized the results established for monotone (k=1) and convex (k=2) densities by giving a characterization of the Maximum Likelihood and Least Square estimators (MLE and LSE) and deriving minimax bounds for rates of convergence. For k strictly larger than 2, the pointwise asymptotic behavior of the MLE and LSE studied by Balabdaoui and Wellner (2007) would show that the MLE and LSE attain the minimax lower bounds in a local pointwise sense. However, the theory assumes that a certain conjecture about the approximation error of a Hermite spline holds true. The main goal of the present note is to show why such a conjecture cannot be true. We also suggest how to bypass the conjecture and rebuild the key proofs in the limit theory of the estimators.
The assumption of log-concavity is a flexible and appealing nonparametric shape constraint in distribution modelling. In this work, we study the log-concave maximum likelihood estimator (MLE) of a probability mass function (pmf). We show that the MLE is strongly consistent and derive its pointwise asymptotic theory under both the well- and misspecified setting. Our asymptotic results are used to calculate confidence intervals for the true log-concave pmf. Both the MLE and the associated confidence intervals may be easily computed using the R package logcondiscr. We illustrate our theoretical results using recent data from the H1N1 pandemic in Ontario, Canada.
In this paper, we describe two computational methods for calculating the cumulative distribution function and the upper quantiles of the maximal difference between a Brownian bridge and its concave majorant. The first method has two different variants that are both based on a Monte Carlo approach, whereas the second uses the Gaver-Stehfest (GS) algorithm for numerical inversion of Laplace transform. If the former method is straightforward to implement, it is very much outperformed by the GS algorithm, which provides a very accurate approximation of the cumulative distribution as well as its upper quantiles. Our numerical work has a direct application in statistics: the maximal difference between a Brownian bridge and its concave majorant arises in connection with a nonparametric test for monotonicity of a density or regression curve on [0, 1]. Our results can be used to construct very accurate rejection region for this test at a given asymptotic level.
In this paper, we consider the problem of finding the Least Squares estimators of two isotonic regression curves $g^\circ_1$ and $g^\circ_2$ under the additional constraint that they are ordered; e.g., $g^\circ_1 \le g^\circ_2$. Given two sets of $n$ data points $y_1, ..., y_n$ and $z_1, >...,z_n$ observed at (the same) design points, the estimates of the true curves are obtained by minimizing the weighted Least Squares criterion $L_2(a, b) = \sum_{j=1}^n (y_j - a_j)^2 w_{1,j}+ \sum_{j=1}^n (z_j - b_j)^2 w_{2,j}$ over the class of pairs of vectors $(a, b) \in \mathbb{R}^n \times \mathbb{R}^n $ such that $a_1 \le a_2 \le ...\le a_n $, $b_1 \le b_2 \le ...\le b_n $, and $a_i \le b_i, i=1, ...,n$. The characterization of the estimators is established. To compute these estimators, we use an iterative projected subgradient algorithm, where the projection is performed with a "generalized" pool-adjacent-violaters algorithm (PAVA), a byproduct of this work. Then, we apply the estimation method to real data from mechanical engineering.
We provide a representation of the maximal difference between a standard Brownian bridge and its concave majorant on the unit interval, from which we deduce expressions for the distribution and density functions and moments of this difference. This maximal difference has an application in nonparametric statistics where it arises in testing monotonicity of a density or regression curve.
We establish limit theory for the Grenander estimator of a monotone density near zero. In particular we consider the situation when the true density $f_0$ is unbounded at zero, with different rates of growth to infinity. In the course of our study we develop new switching relations by use of tools from convex analysis. The theory is applied to a problem involving mixtures.
We find limiting distributions of the nonparametric maximum likelihood estimator (MLE) of a log-concave density, that is, a density of the form $f_0=\expφ_0$ where $φ_0$ is a concave function on $\mathbb{R}$. The pointwise limiting distributions depend on the second and third derivatives at 0 of $H_k$, the "lower invelope" of an integrated Brownian motion process minus a drift term depending on the number of vanishing derivatives of $φ_0=\log f_0$ at the point of interest. We also establish the limiting distribution of the resulting estimator of the mode $M(f_0)$ and establish a new local asymptotic minimax lower bound which shows the optimality of our mode estimator in terms of both rate of convergence and dependence of constants on population values.
We study the asymptotic behavior of the Maximum Likelihood and Least Squares Estimators of a $k$-monotone density $g_0$ at a fixed point $x_0$ when $k>2$. We find that the $j$th derivative of the estimators at $x_0$ converges at the rate $n^{-(k-j)/(2k+1)}$ for $j=0,...,k-1$. The limiting distribution depends on an almost surely uniquely defined stochastic process $H_k$ that stays above (below) the $k$-fold integral of Brownian motion plus a deterministic drift when $k$ is even (odd). Both the MLE and LSE are known to be splines of degree $k-1$ with simple knots. Establishing the order of the random gap $τ_n^+-τ_n^-$, where $τ_n^{\pm}$ denote two successive knots, is a key ingredient of the proof of the main results. We show that this ``gap problem'' can be solved if a conjecture about the upper bound on the error in a particular Hermite interpolation via odd-degree splines holds.
Kiefer and Wolfowitz [Z. Wahrsch. Verw. Gebiete 34 (1976) 73--85] showed that if $F$ is a strictly curved concave distribution function (corresponding to a strictly monotone density $f$), then the Maximum Likelihood Estimator $\hat{F}_n$, which is, in fact, the least concave majorant of the empirical distribution function $\mathbb {F}_n$, differs from the empirical distribution function in the uniform norm by no more than a constant times $(n^{-1}\log n)^{2/3}$ almost surely. We review their result and give an updated version of their proof. We prove a comparable theorem for the class of distribution functions $F$ with convex decreasing densities $f$, but with the maximum likelihood estimator $\hat{F}_n$ of $F$ replaced by the least squares estimator $\widetilde{F}_n$: if $X_1,..., X_n$ are sampled from a distribution function $F$ with strictly convex density $f$, then the least squares estimator $\widetilde{F}_n$ of $F$ and the empirical distribution function $\mathbb {F}_n$ differ in the uniform norm by no more than a constant times $(n^{-1}\log n)^{3/5}$ almost surely. The proofs rely on bounds on the interpolation error for complete spline interpolation due to Hall [J. Approximation Theory 1 (1968) 209--218], Hall and Meyer [J. Approximation Theory 16 (1976) 105--122], building on earlier work by Birkhoff and de Boor [J. Math. Mech. 13 (1964) 827--835]. These results, which are crucial for the developments here, are all nicely summarized and exposited in de Boor [A Practical Guide to Splines (2001) Springer, New York].
We present a Hermite interpolation problem via splines of odd-degree which, to the best knowledge of the authors, has not been considered in the literature on interpolation via odd-degree splines. In this new interpolation problem, we conjecture that the interpolation error is bounded in the supremum norm independently of the locations of the knots. Given an integer k greater than or equal to 3, our spline interpolant is of degree 2k-1 and with 2k-4 (interior) knots. Simulations were performed to check the validity of the conjecture. We present strong numerical evidence in support of the conjecture for k=3, ..., 10 when the interpolated function belongs to C^{(2k)}[0,1], the class of 2k-times continuously differentiable functions on [0,1]. In this case, the worst interpolation error is proved to be attained by the perfect spline of degree 2k with the same knots as the spline interpolant. This interpolation problem arises naturally in nonparametric estimation of a multiply monotone density via Least Squares and Maximum Likehood methods.
Shape constrained densities are encountered in many nonparametric estimation problems. The classes of monotone or convex (and monotone) densities can be viewed as special cases of the classes of k-monotone densities. A density g is said to be k-monotone if its successive derivatives up to order k-2 are alternately negative or positive, with the k-2 derivative convex (or concave) according as k is even (or odd) respectively. g is simply nonincreasing if k=1. These classes of shape constrained densities bridge the gap between the classes of monotone (1-monotone) and convex decreasing (2-monotone) densities for which asymptotic results are known, and the class of completely monotone densities with k infinite. In this series of four papers we consider both (nonparametric) Maximum Likelihood estimators and Least Squares estimators of a k-monotone density. In this first part (part 1), we prove existence of the estimators and give characterizations. We also establish consistency properties, and show that the estimators are splines of order k (degree k-1) with simple knots. We further provide asymptotic minimax risk lower bounds for estimating a k-monotone density g_0 (x_0) and its derivatives g_0^{(j)}(x_0), for j=1, ..., k-1, at a fixed point x_0 under the assumption that the k-th derivative of g_0 at x_0 is non-zero.