Second-order model selection in mixture experiments

Conference · Wed Jul 01 00:00:00 EDT 1992

OSTI ID:10184237

Redgate, P E; Piepel, G F; Hrma, P R

Full second-order models for q-component mixture experiments contain q(q+l)/2 terms, which increases rapidly as q increases. Fitting full second-order models for larger q may involve problems with ill-conditioning and overfitting. These problems can be remedied by transforming the mixture components and/or fitting reduced forms of the full second-order mixture model. Various component transformation and model reduction approaches are discussed. Data from a 10-component nuclear waste glass study are used to illustrate ill-conditioning and overfitting problems that can be encountered when fitting a full second-order mixture model. Component transformation, model term selection, and model evaluation/validation techniques are discussed and illustrated for the waste glass example.

• Research Organization: Pacific Northwest Lab., Richland, WA (United States)
• Sponsoring Organization: USDOE, Washington, DC (United States)
• DOE Contract Number: AC06-76RL01830
• OSTI ID: 10184237
• Report Number(s): PNL-SA-20520; CONF-920893-5; ON: DE92041334
• Resource Relation: Conference: Joint statistical meetings of the American Statistical Association, the Biometric Society and the Institute of Mathematical Statistics,Boston, MA (United States),9-13 Aug 1992; Other Information: PBD: Jul 1992
• Country of Publication: United States
• Language: English

Related Subjects

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE WASTES FROM NUCLEAR FACILITIES
99 GENERAL AND MISCELLANEOUS//MATHEMATICS, COMPUTING, AND INFORMATION SCIENCE
MIXTURES
MAXIMUM-LIKELIHOOD FIT
LOW-LEVEL RADIOACTIVE WASTES
VITRIFICATION
ALPHA-BEARING WASTES
RADIOACTIVE WASTE PROCESSING
STATISTICS
GLASS
SILICON OXIDES
IRON OXIDES
ALUMINIUM OXIDES
SODIUM OXIDES
CALCIUM OXIDES
MAGNESIUM OXIDES
LITHIUM OXIDES
052001
990200
WASTE PROCESSING
MATHEMATICS AND COMPUTERS