Advanced Tools for Mathematics and Statistics Statistical Tools Title: Uncertainty evaluation by Bootstrap: simple statistics and regression analysis Author: Patrizia Ciarlini  Institute: CNR - IAC  Owner: Patrizia Ciarlini  Access level: Document general information: Guest,  Attachments: Guest Abstract: A new computational tool developed at IAC allows interactive bootstrap calculations also on the Internet. Bootstrap methods are a class of estimation procedures quite useful when no reasonable assumptions on the underlying statistical model can be made. They are characterised by an intensive computational approach. The work at IAC has been mainly concentrated on solving the issues involved in providing these tools in the Internet framework. We have devised what we call the Bootstrap Calculator, which is implemented both as a Java applet, running in a browser, and as a stand-alone Java application. Simple statistics are available (mean, weighted mean, median, variance, correlation), but also regression (multiple or polynomial) and several numerical and graphical results are provided.  Created: 5/20/2005 12:55:15 PM Last modified: 5/20/2005 12:55:15 PM Contents: 1 Uncertainty evaluation by Bootstrap (PowerPoint presentation)  (360.00 Kb);  Additional information: View 2 Uncertainty evaluation by Bootstrap (PDF file)  (858.39 Kb);  Additional information: View This work was presented at the International Workshop held in Ljubljana, September 13-15, 2004 (a similar version was also presented at International Conference on Computational and Mathematical Methods in Science and Engineering CMMSE 2004, Uppsala ). The Bootstrap Calculator is here used as applied to metrological data sets to evaluate point or interval uncertainty for the weighted mean estimator and for a regression problem. The numerical and graphical results available with the interactive tool in JAVA, the BC (Bootstrap calculator) tool are shown. For regression, different polynomial models are used. Finally a bootstrap-based estimator to evaluate an "overall uncertainty confidence interval" for the regression model is computed and shown.

If you want more information on this work package, please contact the work package leader Dave Rayner Home   News   General information   Login   Workspace