Statistical Rules of Thumb

Note: Supplemental materials are not guaranteed with Rental or Used book purchases.

ISBN: 9780470144480 | 0470144483
Cover: Paperback
Copyright: 9/2/2008

Buy New

Print on Demand: 2-4 Weeks. This item cannot be cancelled or returned.

$98.71
eBook

Available Instantly

Online: 1825 Days

Downloadable: Lifetime Access

$84.38

Statistical Rules of Thumb, Second Edition compiles simple rules that are widely applicable, robust, and elegant, and each captures key statistical concepts. This handbook provides a framework for considering statistical questions such as sample size and design of experiments. Explaining the justification for each rule, this book conveys the various possibilities that statisticians must think of when designing and conducting a study or analyzing its data. It provides a framework for considering such aspects of statistical work such as: randomness and statistical models; sample size; covariation; epidemiology; environmental studies; designing, conducting, and analyzing studies; words, tables, and graphs; and consulting. New rules of thumb are included such as: Sample size for ratios of means; Very non-significant P-values are very significant; Dichotomize continuous variables for odds ratio analysis; and Correlations need to be substantial to gain advantage in ANCOVA. Some rules have been revised for the new edition, i.e. sample size for relative risk and sample size for percentage change. In addition, the references have been completed updated and expanded. A related website www.vanbelle.org provides additional rules, author presentations and more.

Preface to the Second Edition	p. xiv
Preface to the First Edition	p. xvi
Acronyms	p. xix
The Basics	p. 1
Four Basic Questions	p. 2
Observation is Selection	p. 3
Replicate to Characterize Variability	p. 4
Variability Occurs at Multiple Levels	p. 6
Invalid Selection is the Primary Threat to Valid Inference	p. 6
There is Variation in Strength of Inference	p. 7
Distinguish Randomized and Observational Studies	p. 8
Beware of Linear Models	p. 10
Keep Models As Simple As Possible, But Not More Simple	p. 12
Understand Omnibus Quantities	p. 14
Do Not Multiply Probabilities More Than Necessary	p. 15
Use Two-sided p-Values	p. 16
p-Values for Sample Size, Confidence Intervals for Results	p. 18
At Least Twelve Observations for a Confidence Interval	p. 20
Estimate [plus or minus] Two Standard Errors is Remarkably Robust	p. 22
Know the Unit of the Variable	p. 23
Be Flexible About Scale of Measurement Determining Analysis	p. 24
Be Eclectic and Ecumenical in Inference	p. 25
Sample Size	p. 27
Begin with a Basic Formula for Sample Size-Lehr's Equation	p. 29
Calculating Sample Size Using the Coefficient of Variation	p. 31
No Finite Population Correction for Survey Sample Size	p. 35
Standard Deviation and Sample Range	p. 36
Do Not Formulate a Study Solely in Terms of Effect Size	p. 37
Overlapping Confidence Intervals Do Not Imply Nonsignificance	p. 38
Sample Size Calculation for the Poisson Distribution	p. 40
Sample Size for Poisson With Background Rate	p. 41
Sample Size Calculation for the Binomial Distribution	p. 43
When Unequal Sample Sizes Matter; When They Don't	p. 45
Sample Size With Different Costs for the Two Samples	p. 47
The Rule of Threes for 95% Upper Bounds When There Are No Events	p. 49
Sample Size Calculations Are Determined by the Analysis	p. 50
Observational Studies	p. 53
The Model for an Observational Study is the Sample Survey	p. 54
Large Sample Size Does Not Guarantee Validity	p. 55
Good Observational Studies Are Designed	p. 56
To Establish Cause Effect Requires Longitudinal Data	p. 57
Make Theories Elaborate to Establish Cause and Effect	p. 58
The Hill Guidelines Are a Useful Guide to Show Cause Effect	p. 60
Sensitivity Analyses Assess Model Uncertainty and Missing Data	p. 61
Covariation	p. 65
Assessing and Describing Covariation	p. 67
Don't Summarize Regression Sampling Schemes with Correlation	p. 68
Do Not Correlate Rates or Ratios Indiscriminately	p. 70
Determining Sample Size to Estimate a Correlation	p. 71
Pairing Data is not Always Good	p. 73
Go Beyond Correlation in Drawing Conclusions	p. 75
Agreement As Accuracy, Scale Differential, and Precision	p. 77
Assess Test Reliability by Means of Agreement	p. 80
Range of the Predictor Variable and Regression	p. 82
Measuring Change: Width More Important than Numbers	p. 84
Environmental Studies	p. 87
Begin with the Lognormal Distribution in Environmental Studies	p. 88
Differences Are More Symmetrical	p. 90
Know the Sample Space for Statements of Risk	p. 92
Beware of Pseudoreplication	p. 93
Think Beyond Simple Random Sampling	p. 95
The Size of the Population and Small Effects	p. 96
Models of Small Effects Are Sensitive to Assumptions	p. 97
Distinguish Between Variability and Uncertainty	p. 99
Description of the Database is As Important as Its Data	p. 100
Always Assess the Statistical Basis for an Environmental Standard	p. 101
Measurement of a Standard and Policy	p. 102
Parametric Analyses Make Maximum Use of the Data	p. 104
Confidence, Prediction, and Tolerance Intervals	p. 106
Statistics and Risk Assessment	p. 108
Exposure Assessment is the Weak Link in Assessing Health Effects of Pollutants	p. 110
Assess the Errors in Calibration Due to Inverse Regression	p. 111
Epidemiology	p. 113
Start with the Poisson to Model Incidence or Prevalence	p. 114
The Odds Ratio Approximates the Relative Risk Assuming the Disease is Rare	p. 115
The Number of Events is Crucial in Estimating Sample Sizes	p. 120
Use a Logarithmic Formulation to Calculate Sample Size	p. 122
Take No More than Four or Five Controls per Case	p. 124
Obtain at Least Ten Subjects for Every Variable Investigated	p. 125
Begin with the Exponential Distribution to Model Time to Event	p. 127
Begin with Two Exponentials for Comparing Survival Times	p. 129
Be Wary of Surrogates	p. 131
Prevalence Dominates in Screening Rare Diseases	p. 134
Do Not Dichotomize Unless Absolutely Necessary	p. 138
Additive and Multiplicative Models	p. 139
Evidence-Based Medicine	p. 143
Strength of Evidence	p. 144
Relevance of Information: POEM vs. DOE	p. 148
Begin with Absolute Risk Reduction, then Follow with Relative Risk	p. 149
The Number Needed to Treat (NNT) is Clinically Useful	p. 151
Variability in Response to Treatment Needs to be Considered	p. 153
Safety is the Weak Component of EBM	p. 155
Intent to Treat is the Default Analysis	p. 156
Use Prior Information but not Priors	p. 157
The Four Key Questions for Meta-analysts	p. 158
Design, Conduct, and Analysis	p. 163
Randomization Puts Systematic Effects into the Error Term	p. 163
Blocking is the Key to Reducing Variability	p. 165
Factorial Designs and Joint Effects	p. 166
High-Order Interactions Occur Rarely	p. 168
Balanced Designs Allow Easy Assessment of Joint Effects	p. 170
Analysis Follows Design	p. 171
Independence, Equal Variance, and Normality	p. 173
Plan to Graph the Results of an Analysis	p. 177
Distinguish Between Design Structure and Treatment Structure	p. 179
Make Hierarchical Analyses the Default Analysis	p. 181
Distinguish Between Nested and Crossed Designs-Not Always Easy	p. 182
Plan for Missing Data	p. 184
Address Multiple Comparisons Before Starting the Study	p. 186
Know Properties Preserved When Transforming Units	p. 188
Consider Bootstrapping for Complex Relationships	p. 191
Words, Tables, and Graphs	p. 193
Use Text for a Few Numbers, Tables for Many Numbers, Graphs for Complex Relationships	p. 193
Arrange Information in a Table to Drive Home the Message	p. 195
Always Graph the Data	p. 198
Always Graph Results of an Analysis of Variance	p. 199
Never Use a Pie Chart	p. 203
Bar Graphs Waste Ink; They Don't Illuminate Complex Relationships	p. 204
Stacked Bar Graphs Are Worse Than Bar Graphs	p. 206
Three-Dimensional Bar Graphs Constitute Misdirected Artistry	p. 209
Identify Cross-sectional and Longitudinal Patterns in Longitudinal Data	p. 210
Use Rendering, Manipulation, and Linking in High-Dimensional Data	p. 213
Consulting	p. 217
Session Has Beginning, Middle, and End	p. 218
Ask Questions	p. 219
Make Distinctions	p. 220
Know Yourself, Know the Investigator	p. 222
Tailor Advice to the Level of the Investigator	p. 223
Use Units the Investigator is Comfortable With	p. 224
Agree on Assignment of Responsibilities	p. 226
Any Basic Statistical Computing Package Will Do	p. 227
Ethics Precedes, Guides, and Follows Consultation	p. 228
Be Proactive in Statistical Consulting	p. 230
Use the Web for Reference, Resource, and Education	p. 231
Listen to, and Heed the Advice of Experts in the Field	p. 233
Epilogue	p. 236
References	p. 239
Author Index	p. 255
Topic Index	p. 261
Table of Contents provided by Ingram. All Rights Reserved.

Please wait while the item is added to your bag...

FREE SHIPPING

Amazon no longer offers textbook rentals. We do!

Statistical Rules of Thumb

Summary

Author Biography

Table of Contents

Supplemental Materials