Abstract:We offer in this short report the so-called adaptive functional smoothness estimation in the Hilbert space norm sense in the three classical problems of non-parametrical statistic: regression, density and spectral (density) function measurement (estimation).
What problem does this paper attempt to address?
The main problem that this paper attempts to solve is to provide adaptive estimation methods for the smoothness of functions in three classical problems in non - parametric statistics (regression, density estimation, and spectral density estimation). Specifically, the authors focus on the adaptive function smoothness estimation in the sense of Hilbert space norms.
### Background of the Paper and Problem Statement
In non - parametric statistics, the three common problems are:
1. **Regression Function Estimation**: Given a set of variables $\{y_i\}$, $i = 1,2,\ldots,n$, satisfying the model $y_i = f(x_i)+\xi_i$, where $\xi_i$ are independent and identically distributed random errors.
2. **Density Estimation**: Given a sample $\{\xi_i\}$, $i = 1,2,\ldots,n$, these samples have a common density function $f(x)$.
3. **Spectral Density Estimation**: Given a centered stationary sequence $\{\xi_i\}$ of Gaussian distribution, $i = 1,2,\ldots,n$, whose spectral density is $f(x)$.
### Main Objectives
The main objective of the paper is to provide an adaptive method to estimate the smoothness of these functions. Specifically, the authors hope to make progress in the following aspects:
- **Adaptive Estimation**: Traditional estimation methods usually need to know the smoothness of the function in advance, while the adaptive estimation method can automatically adjust the estimation method to achieve the optimal effect without knowing the specific smoothness of the function.
- **Consistency and Asymptotic Optimality**: Ensure that the proposed estimation method is consistent when the sample size $n$ tends to infinity and is asymptotically optimal.
### Method Overview
The authors propose a projection - based method to estimate the function $f$ and select the optimal number of harmonics $N^*$ by minimizing a certain key variable $A(n,N)$. The specific steps are as follows:
1. **Preliminary Estimation**: Define the key variable $A(n,N)=\frac{N}{n}+\rho(N)$, where $\rho(N)$ represents the smoothness index of the function $f$.
2. **Adaptive Estimation**: Select the optimal number of harmonics $N^*$ by minimizing $A(n,N)$.
3. **Asymptotic Analysis**: Use the Central Limit Theorem (CLT) to conduct asymptotic analysis on the estimator and prove its consistency and asymptotic optimality.
### Conclusion
Through detailed theoretical analysis and numerical simulation, the paper proves the consistency and asymptotic optimality of the proposed adaptive estimation method in regression, density estimation, and spectral density estimation problems. In addition, the authors also discuss how to construct confidence intervals in order to evaluate the reliability of the estimation results in practical applications.
### References
The paper cites a large number of relevant literatures, covering research achievements in multiple fields from non - parametric statistics to probability theory, functional analysis, etc., providing readers with rich background knowledge and theoretical support.