-
Differential Equation-Constrained Local Regression for Data with Sparse Design
Authors:
Chunlei Ge,
W. John Braun
Abstract:
Local polynomial regression of order one or higher often performs poorly in areas with sparse data. In contrast, local constant regression tends to be more robust in these regions, although it is generally the least accurate approach, especially near the boundaries of the data. Incorporating information from differential equations, which may approximately or exactly hold, is one way of extending t…
▽ More
Local polynomial regression of order one or higher often performs poorly in areas with sparse data. In contrast, local constant regression tends to be more robust in these regions, although it is generally the least accurate approach, especially near the boundaries of the data. Incorporating information from differential equations, which may approximately or exactly hold, is one way of extending the sparse design capacity of local constant regression while reducing bias and variance. A nonparametric regression method that exploits first-order differential equations is studied in this paper and applied to noisy mouse tumour growth data. Asymptotic biases and variances of kernel estimators using Taylor polynomials with different degrees are discussed. Model comparison is performed for different estimators through simulation studies under various scenarios that simulate exponential-type growth.
△ Less
Submitted 8 July, 2025;
originally announced July 2025.
-
On Data Sharpening in Nonparametric Autoregressive Models
Authors:
Simon Snyman,
Lengyi Han,
W. John Braun
Abstract:
Data sharpening has been shown to reduce bias in nonparametric regression and density estimation. Its performance on nonlinear first order autoregressive models is studied theoretically and numerically in this paper. Although the asymptotic properties of data sharpening are not as favourable in the presence of serial dependence as in bivariate regression with independent responses, it is still fou…
▽ More
Data sharpening has been shown to reduce bias in nonparametric regression and density estimation. Its performance on nonlinear first order autoregressive models is studied theoretically and numerically in this paper. Although the asymptotic properties of data sharpening are not as favourable in the presence of serial dependence as in bivariate regression with independent responses, it is still found to reduce bias under mild conditions on the autoregression function. Numerical comparisons with the bias reduction method of Cheng et al. (2018) indicate that data sharpening is competitive in this setting.
△ Less
Submitted 12 May, 2025;
originally announced May 2025.
-
Local Quasi-Exponential Growth Models: Kernel Differential Equation Regression and Sparse Data
Authors:
Chunlei Ge,
W. John Braun
Abstract:
Local polynomial regression struggles with several challenges when dealing with sparse data. The difficulty in capturing local features of the underlying function can lead to a potential misrepresentation of the true relationship. Additionally, with limited data points in local neighborhoods, the variance of estimators can increase significantly. Local polynomial regression also requires a substan…
▽ More
Local polynomial regression struggles with several challenges when dealing with sparse data. The difficulty in capturing local features of the underlying function can lead to a potential misrepresentation of the true relationship. Additionally, with limited data points in local neighborhoods, the variance of estimators can increase significantly. Local polynomial regression also requires a substantial amount of data to produce good models, making it less efficient for sparse datasets. This paper employs a differential equation-constrained regression approach, introduced by \citet{ding2014estimation}, for local quasi-exponential growth models. By incorporating first-order differential equations, this method extends the sparse design capacity of local polynomial regression while reducing bias and variance. We discuss the asymptotic biases and variances of kernel estimators using first-degree Taylor polynomials. Model comparisons are conducted using mouse tumor growth data, along with simulation studies under various scenarios that simulate quasi-exponential growth with different noise levels and growth rates.
△ Less
Submitted 30 April, 2025;
originally announced May 2025.
-
Quick and Simple Kernel Differential Equation Regression Estimators for Data with Sparse Design
Authors:
Chunlei Ge,
W. John Braun
Abstract:
Local polynomial regression of order at least one often performs poorly in regions of sparse data. Local constant regression is exceptional in this regard, though it is the least accurate method in general, especially at the boundaries of the data. Incorporating information from differential equations which may approximately or exactly hold is one way of extending the sparse design capacity of loc…
▽ More
Local polynomial regression of order at least one often performs poorly in regions of sparse data. Local constant regression is exceptional in this regard, though it is the least accurate method in general, especially at the boundaries of the data. Incorporating information from differential equations which may approximately or exactly hold is one way of extending the sparse design capacity of local constant regression while reducing bias and variance. A nonparametric regression method that exploits first order differential equations is introduced in this paper and applied to noisy mouse tumour growth data. Asymptotic biases and variances of kernel estimators using Taylor polynomials with different degrees are discussed. Model comparison is performed for different estimators through simulation studies under various scenarios which simulate exponential-type growth.
△ Less
Submitted 14 June, 2024;
originally announced June 2024.
-
Anisotropic local constant smoothing for change-point regression function estimation
Authors:
John R. J. Thompson,
W. John Braun
Abstract:
Understanding forest fire spread in any region of Canada is critical to promoting forest health, and protecting human life and infrastructure. Quantifying fire spread from noisy images, where regions of a fire are separated by change-point boundaries, is critical to faithfully estimating fire spread rates. In this research, we develop a statistically consistent smooth estimator that allows us to d…
▽ More
Understanding forest fire spread in any region of Canada is critical to promoting forest health, and protecting human life and infrastructure. Quantifying fire spread from noisy images, where regions of a fire are separated by change-point boundaries, is critical to faithfully estimating fire spread rates. In this research, we develop a statistically consistent smooth estimator that allows us to denoise fire spread imagery from micro-fire experiments. We develop an anisotropic smoothing method for change-point data that uses estimates of the underlying data generating process to inform smoothing. We show that the anisotropic local constant regression estimator is consistent with convergence rate $O\left(n^{-1/{(q+2)}}\right)$. We demonstrate its effectiveness on simulated one- and two-dimensional change-point data and fire spread imagery from micro-fire experiments.
△ Less
Submitted 7 May, 2024; v1 submitted 30 November, 2020;
originally announced December 2020.
