» » Clinical Versus Statistical Prediction: A Theoretical Analysis and a Review of the Evidence

Download Clinical Versus Statistical Prediction: A Theoretical Analysis and a Review of the Evidence epub

by Paul E. Meehl

This volume explores clinical issues, such as: can we rely on clinical expertise in making decisions about people's lives; when should statistical data be used; and what kind of treatment is best for a particular patient?
Download Clinical Versus Statistical Prediction: A Theoretical Analysis and a Review of the Evidence epub
ISBN: 1568218311
ISBN13: 978-1568218311
Category: Fitness
Subcategory: Psychology & Counseling
Author: Paul E. Meehl
Language: English
Publisher: Jason Aronson, Inc. (May 1, 1996)
Pages: 149 pages
ePUB size: 1760 kb
FB2 size: 1245 kb
Rating: 4.6
Votes: 389
Other Formats: docx txt azw lrf

This is probably one of the most important books ever written (early 50's), and as a result widely ignored by fields that relied (or rely) upon subjective judgements such as medicine. The last time I read anything about this issue, there had not been one study that contradicted Meehl's basic findings (see for example, work by Hastie and Dawes). Yet, irrational decision making is pervasive. Years ago I went through medical school and a psychiatry residency without one reference to this book. I had to find this on my own, and fortunately did so in the late 70's. This and issues such as construct validity, improper linear models, and at least the idea of Bayesian statistics got me into a lot of trouble with my fellow workers in mental health who made totally irrational clinical decisions seriously harming many people (I worked in public mental health centers for the most part); so ultimately Meehl was responsible for my ending a very frustrating career in public psychiatry. Long story. People always tell me I should write a book. Of all the things I studied for many years, Meehl was the foundation.
More than half a century after its original publication in 1954, this remains the most important scholarly work on this subject. As Meehl noted in this monograph and elsewhere, he most emphatically was not comparing different ways of gathering information (e.g., tests vs. interviews). Rather, he was comparing different ways of integrating whatever information has been gathered to make the most valid predictions possible, and thereby to use the most ethically defensible decision-making procedure. One could gather the information through tests (standardized, projective, etc.), interviews (structured, unstructured, or semi-structured), archival records (school, medical, criminal, etc.), informants (family, coworkers, etc.), and a number of other ways.

Meehl's analysis dealt with how all of this information is combined. Should we do it in our heads, or should we use a mechanical decision aid, such as an equation or an actuarial table? He presented a sophisticated analysis of the task and considered conditions that might favor each approach. In one chapter, he reviewed the available empirical data, but this is not the focal point of this monograph.

An article by Grove and Meehl (1996) provides an excellent treatment of objections to mechanical decision aids, with persuasive rebuttals for each, and a meta-analysis by Grove et al (2000) provide a quantitative review of the empirical literature on the subject, including 136 studies that met fairly strict inclusion criteria (in contrast to the 20 studies that were available to Meehl for review in 1954).

Meehl's original conclusion that mechanical decision aids achieve validity rates equal or superior to those of experts exercising their unaided judgment has stood the test of time and is now supported by such a large and diverse array of studies that an impartial reader can reasonably conclude that this "controversy" has been settled.
Despite the ago of this book (written around 1954 as an expansion of lectures given in the years 1947-1950, fifth printing (unchanged) in 1963), the points raised by Paul Meehl remain current.
Paul Meehl asks himself the question what "how can we predict how a person is going to behave?". He distinguishes 2 main approaches: clinical interviews VS statistics (psychometric tests) and discusses the pros and cons of both approaches.
For decades psychologists have been struggling between the use of tests (statistics) and (clinical) interviews. I ran into the problem myself when I did a Whiplash study in 1999 and found that many doctors made clear mistakes during the patient's LAB Profile interviews, to such an extend that the produced data was unreliable for further research! Since then, I reluctantly moved over to the testing side and co-developed the iWAM test (see, but I recommend complementing tests with structured follow up interviews to check the validity of the test answers.
One of his main points pro testing is that "Every hour a clinician spends in thinking and talking about whom to treat, and how and how long, is being subtracted from the available pool of therapeutic time itself." The main counter argument of a clinician will be that every individual is a separate case and thus becomes hard to find in the numbers.
In global, one can say that Meehl holds a quite impartial point of view and tries to present approaches in a factual manner, trying to bring them together. A slight bias towards testing may be expected, given that Paul Meehl was a professor of Psychology and psychiatry at the University of Minnesota and also wrote a book on the clinical use of MMPI.
Complementary books you may what to read are "Psychological testing" by Kaplan & Saccuzzo (2001) and "How to think straight about psychology" by Stanovich, 2001
Good book at a good price!