CMU Artificial Intelligence Repository
RCSTAT: Produces probability intervals from raw data using
the Kyburg/Pollock method.
Few uncertainty reasoning programs go directly from raw observations
to probabilities. Statistical techniques in the Neyman-Pearson
tradition are widely available, but require more expert intervention
than is acceptable for some AI applications. The main problem is
identifying the reference class: as past instances are required to
resemble the target more closely, the sample of such instances
diminishes. The logic for finding the right balance between relevance
and sufficient sample size is not treated in mathematical statistics,
but is treated in the philosophical foundations of probability.
Email from R. Loui.
Requires: C; compile with math library (cc rcstat.c -lm)
Copying: Copyright (c) 1992 by A. Costello, R. Loui
Use, copying, and distribution permitted, provided that
RCSTAT (and derivatives) are not included in a
CD-ROM: Prime Time Freeware for AI, Issue 1-1
Author(s): A. Costello
Contact: Dr. R. P. Loui
Dept. of Computer Science
St. Louis, MO 63130
Authors!Costello, Authors!Loui, C!Code, Data,
Defeasible Reasoning, Induction, Probabilistic Reasoning,
RCSTAT, Reasoning!Defeasible Reasoning,
Reasoning!Statistical Reasoning, Reference Class,
Statistical Reasoning, Statistics, Uncertainty
A user's manual and typescript may be found at the top of the
Other relevant references include:
H. Kyburg, PROBABILITY AND THE LOGIC OF RATIONAL BELIEF, Wesleyan, 1961.
H. Kyburg, LOGICAL FOUNDATIONS OF STATISTICAL INFERENCE, Reidel, 1974.
H. Kyburg, ``The reference class,'' PHILOSOPHY OF SCIENCE 50, 1982.
H. Kyburg, EPISTEMOLOGY AND INFERENCE,Minnesota, 1983.
J. Pollock, ``A theory of direct inference,'' THEORY AND
DECISION 16, 1983.
J. Pollock, ``Foundations for direct inference,'' THEORY AND
DECISION 17, 1984.
J. Pollock, NOMIC PROBABILITY AND THE FOUNDATIONS OF INDUCTION,
R. Loui, ``Computing reference classes,'' UNCERTAINTY IN AI 2, 1987.
Last Web update on Mon Feb 13 10:27:52 1995