Alternative Tests for the Selection of Model Variables (Classic Reprint) - Hardcover

Nathaniel J. Mass

Hardcover

ISBN 10: 0331900319 ISBN 13: 9780331900316

Publisher: Forgotten Books, 2017

Synopsis

Discover how to choose model variables using whole-model testing rather than single-equation statistics.

The work argues that traditional single-equation tests often fail to identify which variables truly shape model behavior. It contrasts these tests with model-behavior analysis, which examines how removing or constraining a variable changes the overall system. Readers will see why data usefulness, not a variable’s isolated significance, should guide variable selection.

Through practical examples at the firm level and in macroeconomic contexts, the material shows how feedback loops and interconnections matter. It demonstrates that a complete system view can clarify when a variable is essential and when its apparent impact is misleading.

• Learn why t-tests and partial correlation coefficients may mislead about a variable’s importance.
• Understand the difference between measuring data usefulness and model specification.
• Explore how whole-model testing analyzes feedback among variables to assess impact.
• See concrete examples at both company and national-economy scales.

Ideal for readers interested in economics, statistics, and the evaluation of social theories through system-wide analysis.

Edition details

• Publisher: Forgotten Books
• Publication date: 2017
• Language: English
• ISBN 10: 0331900319
• ISBN 13: 9780331900316
• Binding: Hardcover
• Number of pages: 113