Comments on: [MADS] plug-in estimator http://hea-www.harvard.edu/AstroStat/slog/2009/mads-plug-in-estimator/ Weaving together Astronomy+Statistics+Computer Science+Engineering+Intrumentation, far beyond the growing borders Fri, 01 Jun 2012 18:47:52 +0000 hourly 1 http://wordpress.org/?v=3.4 By: hlee http://hea-www.harvard.edu/AstroStat/slog/2009/mads-plug-in-estimator/comment-page-1/#comment-888 hlee Tue, 09 Jun 2009 22:35:59 +0000 http://hea-www.harvard.edu/AstroStat/slog/?p=2199#comment-888 I cannot give you a short answer but can say that it is jargon in nonparametric statistics. Before parameterization or setting models (likelihoods and priors from a distribution family), one wishes to characterize a sample distribution by looking its central location and scale (skewness, kurtosis, empirical cdf, contours, principle components, covariance matrix, etc), which are navigated via computing average and sample standard deviation. The equation of computing average, for example, is a plug-in estimator. Since we haven't posed Gaussian or Poisson distribution as the ground truth, we cannot say the average plug-in estimator is the mean (mu in Gaussian or lambda in Poisosn) estimator. We can locate the center of data distribution from a plug-in estimator and thus, without parameterization but based on probability theory and asymptotics, we can quantify the uncertainty of that location plug-in estimator. I cannot give you a short answer but can say that it is jargon in nonparametric statistics. Before parameterization or setting models (likelihoods and priors from a distribution family), one wishes to characterize a sample distribution by looking its central location and scale (skewness, kurtosis, empirical cdf, contours, principle components, covariance matrix, etc), which are navigated via computing average and sample standard deviation. The equation of computing average, for example, is a plug-in estimator. Since we haven’t posed Gaussian or Poisson distribution as the ground truth, we cannot say the average plug-in estimator is the mean (mu in Gaussian or lambda in Poisosn) estimator. We can locate the center of data distribution from a plug-in estimator and thus, without parameterization but based on probability theory and asymptotics, we can quantify the uncertainty of that location plug-in estimator.

]]>
By: yaser http://hea-www.harvard.edu/AstroStat/slog/2009/mads-plug-in-estimator/comment-page-1/#comment-887 yaser Tue, 09 Jun 2009 21:07:01 +0000 http://hea-www.harvard.edu/AstroStat/slog/?p=2199#comment-887 what is plug-in estimator? what is plug-in estimator?

]]>