Implications for model validation of multi-resolution multiperspective modeling (MRMPM) and exploratory analysis / James H. Bigelow and Paul K. Davis.

By:

Bigelow, James H [author.]

Contributor(s):

Publisher: Santa Monica, CA : RAND, 2003Description: xv, 68 pages : illustrations ; 28 cmContent type:

text

Media type:

computer
unmediated

Carrier type:

online resource
volume

ISBN:

0833034820 (pbk.)

Subject(s):

LOC classification:

UG643 .B54 2003

Online resources:

Online Access

Available additional physical forms:

Also available on the internet via WWW in PDF format.

Contents:

Introduction -- Validation-Related Reasons for Multiple Levels of Resolution and Exploratory Analysis -- Consistency and Validation -- Motivated Metamodels, Explanations, and the Importance of a Good Story -- Appendix A: Using a Simple Model to Explain, Extrapolate From, and Provide Face Validity of Complex-Model Results -- Appendix B: Using a Low-Resolution Calculation to Check a High-Resolution Calculation -- Appendix C: Selecting a Good Test Set of Detailed Scenarios -- Appendix D: Illustrations of the Use of Consistency Definitions -- Appendix E: Basing Extrapolation on a Story -- Appendix F: Motivated Metamodels.

Summary: This monograph draws upon a number of the authors past studies to illustrate with concrete examples how multiresolution, multiperspective modeling (MRMPM) and exploratory analysis relate to model validation when the models are not solidly based in settled theory or empirical testing appropriate to the application in question. It is argued that in such cases, the validation process might reasonably assess a model and its associated databases as "valid for exploratory analysis" or "valid, subject to the principal assumptions underlying the model, for exploratory analysis" for a particular context. A model and its data may not be fully valid, but they may still be both useful and good in more-limited ways. It is important that a model being assessed be comprehensible and explainable, and that its data deal effectively with uncertainty, possibly massive uncertainty. Crucial enabling capabilities are provided by multiresolution, multiperspective modeling, including use of families of models and games, and exploratory analysis. These methods are valuable for extrapolating, generalizing, and abstracting from small sets of analyses accomplished with detailed models; for top-down planning; and for providing broad, synoptic assessments of problem areas. They are also important for achieving a deep understanding of problems and communicating insights credibly to others.

Item type:

Average rating: 0.0 (0 votes)

Holdings ( 0 )
Title notes ( 5 )

No physical items for this record

"Project Air Force."

Includes bibliographical references (p. 65-68).

This monograph draws upon a number of the authors past studies to illustrate with concrete examples how multiresolution, multiperspective modeling (MRMPM) and exploratory analysis relate to model validation when the models are not solidly based in settled theory or empirical testing appropriate to the application in question. It is argued that in such cases, the validation process might reasonably assess a model and its associated databases as "valid for exploratory analysis" or "valid, subject to the principal assumptions underlying the model, for exploratory analysis" for a particular context. A model and its data may not be fully valid, but they may still be both useful and good in more-limited ways. It is important that a model being assessed be comprehensible and explainable, and that its data deal effectively with uncertainty, possibly massive uncertainty. Crucial enabling capabilities are provided by multiresolution, multiperspective modeling, including use of families of models and games, and exploratory analysis. These methods are valuable for extrapolating, generalizing, and abstracting from small sets of analyses accomplished with detailed models; for top-down planning; and for providing broad, synoptic assessments of problem areas. They are also important for achieving a deep understanding of problems and communicating insights credibly to others.

Also available on the internet via WWW in PDF format.