Conservative confidence bands in curvilinear regression (Q1085547)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Conservative confidence bands in curvilinear regression
scientific article

    Statements

    Conservative confidence bands in curvilinear regression (English)
    0 references
    0 references
    1986
    0 references
    Consider the regression model of full rank in which we observe \[ Y_ i=\sum^{k}_{j=1}b_ jf_ j(x_ i)+e_ i,\quad i=1,...,n, \] where \(b_ j\) are unknown, the \(f_ j\) are known, the points \(x_ i\) are known, the \(e_ i\) are i.i.d. \(N(0,\sigma^ 2)\). Let \(I\subset {\mathbb{R}}\) be a closed interval, \(f(x)=(f_ 1(x),...,f_ k(x))'\) defines a function from I to \({\mathbb{R}}^ k\), \^b\(\sim N_ k(b,\sigma^ 2\Sigma)\) and \(s^ 2\) are the least squares estimator of \(b=(b_ 1,...,b_ k)'\) and an unbiased estimator of \(\sigma^ 2\) respectively. The paper gives a method for constructing a simultaneous confidence band for the regression function b'f(x) for \(x\in I\). The author considers Scheffe-type bands, that is, bands of the form \^b'f(x)\(\pm csp(x)\) for \(x\in I\), where \(p(x)=\{f(x)'\Sigma f(x)\}^{1/2}\) and \(c\geq 0\). The method is based on a geometric inequality giving an upper bound for the uniform measure of the set of piecewise differentiable paths with finite length in \(S^{k-1}\), the unit sphere in \({\mathbb{R}}^ k\).
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    curvilinear regression
    0 references
    simultaneous confidence band
    0 references
    regression function
    0 references
    Scheffe-type bands
    0 references
    geometric inequality
    0 references
    0 references