Efficiency of a Linear Combination of the Median and the Sample Mean: The Double Truncated Normal Distribution (Q4677119)

This paper deals with statistical models defined on \(R\) with a density function of the form \(f(x-\mu;\psi)\), where \(\mu\in R\) is a location parameter, \(\psi\in\Psi\subset R^{k}\) is a \(k\)-dimensional parameter and \(f(t;\psi)=f(-t;\psi)\) for all \(\psi\in\Psi\). The following result is proved. The symmetric location models which have an asymptotically efficient estimator of the location parameter \(\mu\) of the form \(\bar\mu=w\bar x+(1-w)\tilde x\), where \(\bar x\) and \(\tilde x\) are sample mean and sample median, respectively, have a density function of the form \[ f_{\theta}(x;\mu,\sigma)=[ \phi(\theta)/(2(1-\Phi(\theta))\sigma)]\exp (-\theta(| x-\mu|/\sigma)-(x-\mu)^2/(2\sigma^2)), \] where \(\phi(\theta)\) and \(\Phi(\theta)\) denote the standard normal density and cumulative distribution functions, respectively; \(\mu\in R\) and \(\sigma\in R^{+}\) are the location and scale parameters, and \(\theta\) is a fixed value. Moreover, \(w=w(\theta)=(1-\Phi(\theta))/(1-\Phi(\theta)+\theta\phi(\theta))\). A simple method to estimate the parameters and applications to currency exchange rates are presented.