Statistical Reasoning in Medicine : The Intuitive P Value Primer

by Moye, Lemuel A.

• ISBN13: 9780387989334
• ISBN10: 0387989331
• Edition: 1st
• Format: Paperback
• Copyright: 2000-06-01
• Publisher: Springer Verlag

Summary

Through careful, deliberate thought this book provides the nonmathematician with a foundation for understanding the underlying statistical reasoning process in clinical research. This volume recognizes the inevitable tension between the mathematics of hypothesis testing and the ethical requirements in medical research and concentrates on the resolution of these issues in P value interpretation. By presenting the material in a conversational format and consciously de-emphasizing computational devices and focusing instead on the features of experimental design that either clarify or blur P value interpretation statistical reasoning becomes extremely accessible to the uninitiated.

Table of Contents

Preface	v
Acknowledgments	ix
Introduction	xvii
Prologue: ``Let Others Thrash it out!'' A Brief History	1	(266)
Europe's emergence from the Middle Ages
Intellectual Triumph: The Industrial Revolution
The Rise of Political Arithmetic
The Role of Religion in Political Arithmetic
``Let Others Thrash It Out!''
Early Experimental Design
Fisher, Student, and the Modern Era of Experimental Design
Conclusions
References
Patients, Patience, and P Values	20	(28)
The Dilemma of Physicians
The Great Compromise
Random sampling
Example of sampling error
Fleas
Being Dealt a Bad Hand
Leaky Protocols and Sampling Error Contamination
Study Concordance vs. Study Discordance
Importance of prospective experimental design
Physician-scientist Community Responsibility
Patience and tight protocols
Discovering a New World
Conclusions
Shrine Worship	48	(22)
The Common Nightmare
P Value Bashing
Is War Too Important to Be Left to the Generals?
Finding a Way out
Less than the Sum of Its Parts
Brief review of significance testing
Responsibilities and Mandates
Is there an alternative to p values?
Conclusions
Mistaken Identity: P Values in Epidemiology	70	(27)
Mistaken Identity
Detective Work
Definition of Causation
Determining Causation
Experimental Vs. Observational Studies
Randomization
Collisions
Strengths and Weaknesses of Different Research Designs as Contributors to the Tenets
Tenets of Causality
Strength of Association
Temporality
Experimental Evidence
Analogy
Epidemiologic Research Designs and Problematic P Values
Directionality: Looking Forward or Looking Backward
Timing---Retrospective vs. Prospective
Case Control Studies
Adjustment of Variables in Observational Studies
Design Considerations
Solid Structures from Imperfect Bricks
How to Draw Inferences From Epidemiologic Studies.
Study counting: The ceteris parabus fallacy
Critiquing Experimental Designs
Conclusions
References
Loud Messengers--P Values and Effect Size	97	(23)
P values and Effect sizes---the LRC Trial
Measures of Effect Size
Confidence Intervals
Backdoor Significance Testing
Clinical Significance Without P Values
The Radium Girls
The Buy-In
Do Blood Pressures Lie? The Frye Test
Relative Risk and the Supreme Court
P Values, Confidence Intervals and the U. S. Supreme Court
Conclusions
``Your Searchlight's Not On!'' Power and P Values	120	(19)
The Critic's View
No Way Out?
Sample Size Computations
Continuous Outcome Measures
Dichotamous Outcome Measures
The Right Sample Size---Lipid Research Clinics
The ``Good Enough for Them'' Sample Size Approach
Efficacy Seduction
However, there is a power dividend
References
``Sir, you are unethical!'' One-tailed vs. Two-tailed Testing	139	(20)
Evidence Strength vs. Belief Strength
UMPTs
The arrhythmia suppression hypothesis and CAST
Lipid Research Clinics Results
Sample Size Issues
Hoping for the Best, Preparing for the Worst
Symmetrics vs. Ethics
Conclusion
References
P Values and Multiple Endpoints I: Pernicious Abstractions?	159	(14)
The great trap: experimental efficiency vs. interpretative parsimony
Don't Ask, Don't Tell...
Community Expectations vs. Mandate Abrogation
Let the Investigators Decide After the Fact?
Throwing the Baby Out with the Bath Water? Negative Trials with Positive Secondary Endpoints
Conclusions
References
P values and Multiple Endpoints II: Noxious Placebos in the Population	173	(38)
Noxious Placebo Protection
PA3 S---The Prospective Asymmetric Alpha Allocation System
Some Interim Comments on PA3S
Notation
The Sample Size Straightjacket
An Ethical Move to Asymmetry
New Therapy for Hypercholesterolemia
Heart Failure
Arrhythmias
Type I error Dependency
Alpha Eggs in the Endpoint Basket
Conclusions
References
Neurons vs. Silicon: Regression Analysis and P values	211	(23)
The New Meaning of Regression
Underlying Assumptions in Simple Regression Analysis
Simple Examples and Explanations
Variance Partitioning
Enter Dichotomous Independent Variables
Multiple Regression and the Meaning of ``Adjustment''
Superfits
Pearls of Great Price
Effect Modifiers and Alpha Allocation
Effect Modifiers can be Difficult to Explain.
Conclusion
Bayesian P Values	234	(17)
The Frequentists vs. the Bayesians
The Bayesian Philosophy
Example of Bayesian Estimation
Bayesian P values
Example of Bayes Hypothesis Testing: Asthma Prevalence in a Clinic
Conclusion
References
Blind Guides for Explorers: P Values, Subgroups and Data Dredging	251	(16)
Unsuspected Pots of Gold
Confirmatory vs. Exploratory Analyses
Each Diminishes the Whole
How to Interpret Exploratory Analyses
Caveat emptor: Let the Buyer Beware.
Conclusions
Conclusions: Good Servants but Bad Masters	267	(7)
Index	274