Semiparametric smoothing splines
The article deals with the statistical model \(y_{i}=m(t_{i}) +\varepsilon_{i}\), \(i=1,\ldots,n\), with independent random variables \(\varepsilon_{i}\), \(E\varepsilon_{i}=0\), \(E\varepsilon_{i}^2=\sigma_{i}^2\) and \(m(t)\in W_2^{(l)}[0,1]\), where \(W_2^{(l)}[0,1]\) is a functional space of smooth functions defined as follows: \[ W_2^{(l)}[0,1]= \bigl\{ m : m^{1},\ldots,m^{(l-1)} \text{absolutely continuous and} \int\limits_0^1(m^{(l)}(t))^2 dt <\infty\bigr\}. \] Information concerning the underlying regression function is represented by known smooth functions \(g_{0i}=g_0(t_{i})\), \(i=1,\ldots,n\), \(g_0(\cdot)\in W_2^{(k)}.\) The author proposes the following estimator of the function \(m(\cdot).\) Let \[ L_{n}(m)={1\over n}\sum\limits_{i=1}^{n} w_{in}(y_{i}- m(t_{i}))^2 +\lambda_1\int\limits_0^1(m^{(l)}(t))^2 +{\lambda_2\over n}\sum\limits_{i=1}^{n}(m(t_{i})- g_0(t_{i}))^2, \] where \(w_{in}={1\over \sigma_{i}}\bigl({1\over n}\sum\limits _{i=1}^{n}{1\over \sigma_{i}^2}\bigr)^{-{1\over 2}}.\) The two parameters \(\lambda_1,\lambda_2\) represent a tradeoff in the requirements on fidelity to data, degree of smoothness and ``closeness to some given function \(g_0(t)\). The estimator satisfies the following optimization problem: \({\hat m}_{\lambda}=arg \min\limits_{m\in W_2^{(l)}[0,1]}L_{n}(m).\) Conditions of existence and uniqueness of such estimators are found. Let \[ R_{n}(\lambda_1,\lambda_2)=n^{-1}\sum\limits_{i=1} ^{n}E(m(t_{i})-{\hat m}_{\lambda}(t_{i}))^2 \] be the risk function. The author investigates the asymptotic properties of \(R_{n}(\lambda_1,\lambda_2)\) as \(n\to \infty.\) Some examples are also considered.
- Estimating effort function with semiparametric model
- Smoothing splines with varying smoothing parameter
- Smoothing parameter selection for a class of semiparametric linear models
- scientific article; zbMATH DE number 3936263 (Why is no real title available?)
- Smoothing spline growth curves with covariates
- SEMIPARAMETRIC REGRESSION WITH MULTIPLICATIVE ADJUSTMENT
- Asymptotic normality of spline estimator when the errors are a linear stationary process
- Fitting smoothing splines to data from multiple sources
- scientific article; zbMATH DE number 7300702 (Why is no real title available?)
- Semiparametric Penalized Spline Regression
- Smoothing Spline Semiparametric Density Models
- Application of structural risk minimization to multivariate smoothing spline regression estimates
- The data-smoothing aspect of Stein estimates
- The average posterior variance of a smoothing spline and a consistent estimate of the average squared error
This page was built for publication: Semiparametric smoothing splines
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4231203)
