No label defined (Q3749907)

From MaRDI portal

!

This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:

scientific article; zbMATH DE number 3986385
Language Label Description Also known as
default for all languages
No label defined
    English
    No label defined
    scientific article; zbMATH DE number 3986385

      Statements

      1986
      0 references
      rate of convergence
      0 references
      smoothness
      0 references
      Fourier series
      0 references
      density estimation
      0 references
      Fourier coefficient
      0 references
      orthogonal series
      0 references
      square-integrable density
      0 references
      expansion
      0 references
      integrated mean square error
      0 references
      asymptotically optimal
      0 references
      lower bound
      0 references
      0 references
      The nonparametric estimation of a density f with compact support \(A\subset R^ d\) is considered, and it is shown that a lower bound to the rate of convergence is achieved by estimators based on orthogonal series. Let \(\| A\|\) denote the content of A and let \(\{\phi_ i\}\) be a complete, orthogonal sequence on A. A square-integrable density f on A admits an \(L^ 2\)-convergent expansion, \(f\sim \sum c_ i\phi_ i\), where \(\hat c{}_ i=\int \phi_ i(X_ j)\) is an unbiased estimate of \(c_ i\). An estimator of f is defined by \[ (1)\;\hat f(.| b)=\sum b_ i\hat c_ i\phi_ i \] where \(\sum | b_ i| <\infty\) is usually assumed for the sequence \(b=(b_ 1,b_ 2,...)\) to ensure convergence of the random series in (1). It is shown that the integrated mean square error \[ (2)\;\sup_{f\in F(m)}\int_{A}E(\tilde f(x)- f(x))^ 2dx\geq C_ 0\min (1,\epsilon)\sum^{\infty}_{i=m}\min (n^{-1},a^ 2_ i), \] where \(\tilde f\) is any square-integrable, nonparametric estimator of f (based on a random sample of size n), F(m) is the class of all f of the form \[ (3)\quad f(x)=\| A\|^{- 1}+\sum^{\infty}_{i=m}r_ ia_ i\phi_ i(x), \] and the constants \(r_ i\), \(a_ i\), and \(\epsilon\) are restricted to ensure that f is a density. It is also shown that for certain choices of \(b\equiv b_ n\), the estimators of the form (1) are asymptotically optimal in achieving the lower bound in (2).
      0 references

      Identifiers