Adaptive wavelet estimator for nonparametric density deconvolution. (English) Zbl 0962.62030

From the introduction: Let \(\theta\) and \(\varepsilon\) be independent random variables with density functions \(g\) and \(q\), respectively, where \(g\) is unknown and \(q\) is known. One observes a sample of random variables, \[ X_i=\theta_i +\varepsilon_i, \quad i=1,2, \dots,n.\tag{1} \] The objective is to estimate the density function \(g\). In this situation the density function \(p\) of \(X_i\), \(i=1,\dots,n\), is the convolution of \(q\) and \(g\), \[ p(x)= \int^\infty_{-\infty} q(x-\theta) g(\theta)d \theta. \tag{2} \] Hence the problem of estimating \(g\) in (2) is called a deconvolution problem. The problem arises in many applications and, therefore, it was studied extensively in the last decade. The most popular approach to the problem was to estimate \(p(x)\) by a kernel estimator and then solve equation (2) using a Fourier transform.
This paper deals with the estimation of a deconvolution density using a wavelet decomposition. The underlying idea is to present \(g(\theta)\) via a wavelet expansion and then to estimate the coefficients using a deconvolution algorithm. The proposed approach is based on orthogonal series methods for the estimation of a prior density, and also on modern developments of wavelet techniques in curve estimation. The estimators proposed in this paper are based on Meyer-type wavelets rather than on wavelets with bounded support. Meyer-type wavelets form a subset of the set of band-limited wavelets that allow immediate deconvolution. It should be noted that the nonlinear wavelet estimator constructed in this paper is based on a “global thresholding” which is somewhat different from the “block thresholding”: in the “global thresholding” procedure all coefficients of the same level are thresholded simultaneously, while “block thresholding” groups together only a finite number of coefficients.
This article is organized in the following way. In Section 2 we give a brief description of Meyer-type wavelets and derive the linear and nonlinear wavelet estimators of \(g(\theta)\). In Section 3 we investigate asymptotic behavior of the estimators when \(g(\theta)\in H^\alpha\). The case of supersmooth \(g(\theta)\) is considered in Section 4. In Section 5 we illustrate the theory by examples. Section 6 concludes the paper with discussion. Section 7 contains proofs of the theorems.


62G07 Density estimation
62G05 Nonparametric estimation
62G20 Asymptotic properties of nonparametric inference
42C40 Nontrigonometric harmonic analysis involving wavelets and other special systems
Full Text: DOI


[1] Abramovich, F. and Silverman, B. W. (1998). Wavelet decomposition approaches to statistical inverse problems. Biometrika 85 115-129. JSTOR: · Zbl 0908.62095
[2] Antoniadis, A., Grégoire, G. and McKeague, I. W. (1994). Wavelet method for curve estimation. J. Amer. Statist. Assoc. 89 1340-1353. JSTOR: · Zbl 0815.62018
[3] Carroll, R. J. and Hall, P. (1988). Optimal rates of convergence for deconvolving a density. J. Amer. Statist. Assoc. 83 1184-1186. JSTOR: · Zbl 0673.62033
[4] Desouza, C. M. (1991). An empirical Bayes formulation of cohort models in cancer epidemiology. Statistics in Medicine 10 1241-1256.
[5] Devroye, L. (1989). Consistent deconvolution in density estimation. Canad. J. Statist. 17 235- 239. JSTOR: · Zbl 0679.62029
[6] Diggle, P. J. and Hall, P. (1993). A Fourier approach to nonparametric deconvolution of a density estimate. J. Roy. Statist. Assoc. Ser. B 55 523-531. JSTOR: · Zbl 0783.62030
[7] Donoho, D. and Johnstone, I. (1995). Adapting to unknown smoothness via wavelet shrinkage. J. Amer. Statist. Assoc. 90 1200-1224. JSTOR: · Zbl 0869.62024
[8] Donoho, D., Johnstone, I., Kerkyacharian, G. and Picard, D. (1996). Density estimation by wavelet thresholding. Ann. Statist. 24 508-539. · Zbl 0860.62032
[9] Efromovich, S. (1997). Density estimation for the case of supersmooth measurement error. J. Amer. Statist. Assoc. 92 526-535. Fan, J. (1991a). On the optimal rates of convergence for nonparametric deconvolution problem. Ann. Statist. 19 1257-1272. Fan, J. (1991b). Asymptoticnormality for deconvolution kernel density estimators. Sankhy \?a Ser. A 53 97-110. Fan, J. (1991c). Global behavior of deconvolution kernel estimates. Statist. Sinica 1 541-551. · Zbl 0890.62027
[10] Fan, J. (1993). Adaptively local one-dimensional subproblems with application to a deconvolution problem. Ann. Statist. 21 600-610. · Zbl 0785.62038
[11] Hall, P., Kerkyacharian, G. and Picard, D. (1998). Block threshold rules for curve estimation using kernel and wavelet methods. Ann. Statist. 26 922-942. · Zbl 0929.62040
[12] Hall, P. and Patil, P. (1995). Formulae for mean integrated squared error of nonlinear waveletbased density estimators. Ann. Statist. 23 905-928. · Zbl 0839.62042
[13] Hall, P., Penev, S., Kerkyacharian, G. and Picard, D. (1997). Numerical performance of block thresholded wavelet estimators. Statistics Comput. 7 115-124.
[14] Hernández, E. and Weiss, G. (1996). A First Course on Wavelets. CRC Press, Boca Raton, FL. · Zbl 0885.42018
[15] Kerkyacharian, G. and Picard, D. (1992). Density estimation in Besov spaces. Statist. Probab. Lett. 13 15-24. · Zbl 0749.62026
[16] Liu, M. C. and Taylor, R. L. (1989). A consistent nonparametric density estimator for the deconvolution problem. Canad. J. Statist. 17 427-438. JSTOR: · Zbl 0694.62017
[17] Louis, T. A. (1991). Using empirical Bayes methods in biopharmaceutical research. Statistics in Medicine 10 811-827.
[18] Masry, E. (1991). Multivariate probability density deconvolution for stationary random processes. IEEE Trans. Inform. Theory 37 1105-1115. Masry, E. (1993a). Strong consistency and rates for deconvolution of multivariate densities of stationary processes. Stochastic Process. Appl. 47 53-74. Masry, E. (1993b). Asymptoticnormality for deconvolution estimators of multivariate densities of stationary processes. J. Multivariate Anal. 44 47-68. · Zbl 0732.60045
[19] Masry, E. (1994). Probability density estimation from dependent observations using wavelet orthonormal bases. Statist. Probab. Lett. 21 181-194. · Zbl 0814.62021
[20] Penskaya, M. (1985). Projection estimators of the density of an a priori distribution and of functionals of it. Theory Probab. Math. Statist. 31 113-124. · Zbl 0599.62046
[21] Pensky, M. (1999). Estimation of a smooth density function using Meyer-type wavelets. Statist. Decisions 17 111-123 · Zbl 0941.62042
[22] Piterbarg, V. and Penskaya, M. (1993). On asymptoticdistribution of integrated squared error of an estimate of a component of a convolution. Math. Methods Statist. 2 30-41. · Zbl 0798.62026
[23] Stefansky, L. A. (1990). Rates of convergence of some estimators in a class of deconvolution problems. Statist. Probab. Lett. 9 229-235. · Zbl 0686.62026
[24] Stefanski, L. and Carrol, R. J. (1990). Deconvoluting kernel density estimators. Statistics 21 169-184. · Zbl 0697.62035
[25] Taylor, R. L. and Zhang, H. M. (1990). On a strongly consistent non-parametric density estimator for deconvolution problem. Comm. Statist. Theory Methods 19 3325-3342. · Zbl 0734.62043
[26] Talagrand M. (1994). Sharper bounds for empirical processes. Ann. Probab. 22 28-76. · Zbl 0798.60051
[27] Walter, G. G. (1981). Orthogonal series estimators of the prior distribution. Sankhy\?a Ser. A 43 228-245. · Zbl 0528.62025
[28] Walter, G. G. (1994). Wavelets and Other Orthogonal Systems with Applications. CRC Press, Boca Raton, FL. · Zbl 0866.42022
[29] Zayed, A. I. and Walter, G. G. (1996). Characterization of analytic functions in terms of their wavelet coefficients. Complex Variables 29 265-276. · Zbl 0852.30004
[30] Zhang, C. H. (1990). Fourier methods for estimating mixing densities and distributions. Ann. Statist. 18 806-831. · Zbl 0778.62037
[31] Zhang, C. H. (1992). On deconvolution using time of flight information in positron emission tomography. Statist. Sinica 2 553-575. · Zbl 0822.62029
This reference list is based on information provided by the publisher or from digital mathematics libraries. Its items are heuristically matched to zbMATH identifiers and may contain data conversion errors. It attempts to reflect the references listed in the original paper as accurately as possible without claiming the completeness or perfect precision of the matching.