A Semiparametric Approach to Density Estimation

From MaRDI portal
Publication:3769773





One method of fitting a parametric density function \(f(x,\theta)\) is first to estimate \(\theta\) by maximum likelihood, say, and then to estimate \(f(x,\theta)\) by \(f(x,{\hat\theta})\). On the other hand, when the parametric model does not hold, the true density f(x) may be estimated nonparametrically, as in the case of a kernel estimate \(\hat f(x)\). The key idea proposed is to fit a combination of the parametric and nonparametric estimates, namely, \[ g(x,\pi)=\pi f(x,{\hat \theta})+(1-\pi)\hat f(x), \] where \(\pi\) \((0\leq \pi \leq 1)\) is unknown. The parameter \(\pi\) is estimated from the data, and its estimate \({\hat\pi}\) is then used in \(g(x,{\hat\pi})\) as the proposed density estimate. The main point is that we expect \({\hat\pi}\) to be close to unity when the parametric model prevails, and close to zero when it does not hold. We show that, under certain conditions, \(g(x,{\hat\pi}\)) converges to the parametric density when the parametric model holds and approaches the true \(f(x)\) when the parametric model does not hold. The procedure was applied to a number of actual data sets. In each case the maximum likelihood estimate \({\hat\pi}\) was readily obtained and the semiparametric density estimate provided a better understanding of which model prevailed. Other compromise estimates may also be used. One such is the geometric mean of cumulative distribution functions: \[ H(x,\pi)= F^{\pi}(x,{\hat\theta})\hat F^{1-\pi}(x), \] where \(F(x,{\hat\theta})\) is a parametric estimate of the parametric distribution function \(F(x,\theta)\) and \(\hat F(x)\) is a nonparametric estimate of \(F(x)\).




Cited in
(57)








This page was built for publication: A Semiparametric Approach to Density Estimation

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3769773)