MPR is a stochastic method in statistical estimation that gains the sample information from the point arrangement along the model curve. The calculus for this ‘longitudinal advancing’ foots on the theory of runs (iterations). Objective function is the pdf of the items leading to a maximum of the methods specific run-symmetry. The solution appears as a data scope dependent interval of equivalent parameters. The method is not sensitive to the data error distribution and copes with non-linearity in the model parameters. MPR can substitute common hypothesis testing. Optimality is postulated by a transcendent presentation of evidence and by a stochastic isomorphy. MPR shows behaviour patterns known from quantum mechanics.- Drawback is that the scopes of the observation samples must be modest; this limitation is forced by the maximum integer value (i.e. the internal bus size) of the machine – a way out we propose on the basis of the found fundamental iteration-pdf .
Keywords: Cluster, error distribution, estimation interval, frequency classification, genome, histogram, hypothesis testing, non-linear 'regression', mixture, non-normal, nonparametric, outlier, quantile regression, robustness, residual
|Download:||ABOUT THE NOVEL MPR SYMMETRATION ESTIMATOR (PDF, 2,5 MB)|
Herausgeber und Autor: Bernhard R. Markowski · Bahnhofstrasse 72 · D-36341 Lauterbach
Überarbeitete Version Mai 2017 / revised version may 2017