Subscribe Now Subscribe Today
Science Alert
FOLLOW US:     Facebook     Twitter
Curve Top
Asian Journal of Mathematics & Statistics
  Year: 2015 | Volume: 8 | Issue: 1 | Page No.: 19-34
DOI: 10.3923/ajms.2015.19.34
Facebook Twitter Digg Reddit Linkedin StumbleUpon E-mail
On the Adequacy of Variable Selection Techniques on Model Building
M.P. Iwundu and O.P. Efezino

The adequacy of variable selection techniques on model building is examined using the drying characteristics of fresh grains. Three selection techniques (forward selection, backward elimination and stepwise methods) are employed each of which identifies a series of models assumed to be adequate. From the resulting models for each technique, the most suitable model is determined using standard assessment criteria namely, R2, R2adj, PRESS, AIC and Cp-statistic. In addition to the standard assessment criteria, the D-optimality criterion is examined and presented as a criterion for measuring the goodness and adequacy of regression models. Results show that under the forward selection and stepwise regression methods, each assessment criterion locates the same model. Variation seems to exist using backward elimination technique.
PDF Fulltext XML References Citation Report Citation
How to cite this article:

M.P. Iwundu and O.P. Efezino, 2015. On the Adequacy of Variable Selection Techniques on Model Building. Asian Journal of Mathematics & Statistics, 8: 19-34.

DOI: 10.3923/ajms.2015.19.34








Curve Bottom