Accepting Error to Make Less Error.
Abstract
It is argued that the clinical and statistical approaches rest on different assumptions about the nature of random error and the appropriate level of accuracy to be expected in prediction. To examine this, a case is made for each approach. The clinical approach is characterized as being deterministic, causal, and less concerned with prediction than with diagnosis and treatment. The statistical approach accepts error as inevitable and in so doing, makes less error in prediction, This is illustrated using examples from probability learning and equal weighting in linear models. Thereafter, a decision analysis of the two approaches is proposed. Of particular importance are the errors that characterize approach; myths, magic, and illusions of control in the clinical; lost opportunities and illusions of the lack of control in the statistical. It is concluded that each approach represents a gamble with corresponding risks and benefits. Keywords include: Clinical vs. statistical, Prediction, Randomness, and Decision making.
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 01, 1985
- Accession Number
- ADA153949
Entities
People
- H. J. Einhorn
Organizations
- University of Chicago