On the best predictive general linear model for data analysis: A tolerance region algorithm for prediction

C. P. Kitsos, Vasilios Zarikas

Research output: Contribution to journalArticlepeer-review

4 Citations (Scopus)

Abstract

There is a constant need for correct and meaningful statistical prediction. The General Linear Model (GLM) is a commonly used method to fit the data although most of the times the target is to construct a linear model in order to "predict" the value of the dependent variable; a goal for which GLM has not been designed for. The aim of the present study is to work on best model for a future observation, adopting the tolerance regions concept. A new method is explained and demonstrated, which is an alternative approach for choosing the optimal order of a response polynomial. The present study proposes a novel algorithm, which selects the best response polynomial, as far as prediction is concerned. The beta expected tolerance region is applied. The proposed computational approach has been applied for several data sets. This analysis, confirms the utility and the advantage of the method which provides non trivial results.

Original languageEnglish
Pages (from-to)513-524
Number of pages12
JournalJournal of Applied Sciences
Volume13
Issue number4
DOIs
Publication statusPublished - 2013

Keywords

  • Algorithms
  • Case studies
  • General linear regression
  • Predictive models
  • Tolerance intervals

ASJC Scopus subject areas

  • General

Fingerprint Dive into the research topics of 'On the best predictive general linear model for data analysis: A tolerance region algorithm for prediction'. Together they form a unique fingerprint.

Cite this