- •Foreword
- •Contents
- •Contributor Current and Past Positions: Association for Academic Surgery
- •Contributors
- •Academic Surgeons as Bridge-Tenders
- •Types of Surgical Research
- •Going Forward
- •Selected Readings
- •Introduction
- •Preparation Phase
- •Assistant Professor
- •Job Search
- •The First Three Years
- •Career Development Awards (CDAs)
- •Contemplating a Mid-Career Move?
- •Approaching Promotion
- •Associate Professor and Transition to Full Professor
- •Conclusion
- •Selected Readings
- •Introduction
- •Reviewing the Literature
- •Developing a Hypothesis
- •Study Design
- •Selected Readings
- •Introduction
- •The Dual Loyalties of the Surgeon-Scientist
- •Human Subjects Research
- •Informed Consent
- •Surgical Innovation and Surgical Research
- •Conflict of Interest
- •Publication and Authorship
- •Conclusion
- •References
- •Sources of Error in Medical Research
- •Study Design
- •Inferential Statistics
- •Types of Variables
- •Measures of Central Tendency and Spread
- •Measures of Spread
- •Comparison of Numeric Variables
- •Comparison of Categorical Values
- •Outcomes/Health Services Research
- •Steps in Outcomes Research
- •The Basics of Advanced Statistical Analysis
- •Multivariate Analysis
- •Time-to-Event Analysis
- •Advanced Methods for Controlling for Selection Bias
- •Propensity Score Analysis
- •Instrumental Variable (IV) Analysis
- •Summary
- •Selected Readings
- •Transgenic Models
- •Xenograft Models
- •Noncancer Models
- •Alternative Vertebrate Models
- •Selected Readings
- •Overview
- •Intellectual Disciplines and Research Tools
- •Comparative Effectiveness Research
- •Patient-Centered Outcomes Research
- •Data Synthesis
- •Overview
- •Intellectual Disciplines and Research Tools
- •Disparities
- •Quality Measurement
- •Implementation Science
- •Patient Safety
- •Optimizing the Health Care Delivery System
- •Overview
- •Intellectual Disciplines and Research Tools
- •Policy Evaluation
- •Surgical Workforce
- •Conclusion
- •References
- •Introduction
- •What Is Evidence-Based Medicine?
- •Evidence-Based Educational Research
- •Forums for Surgical Education Research
- •Conducting Surgical Education Research
- •Developing Good Research Questions
- •Beginning the Study Design Process
- •Developing a Research Team
- •Pilot Testing
- •Demonstrating Reliability and Validity
- •Developing a Study Design
- •Data Collection and Analysis
- •Surveys
- •Ethics
- •Funding
- •Conclusions
- •Selected Readings
- •Genomics
- •Gene-Expression Profiling
- •Proteomics
- •Metabolomics
- •Conclusions
- •References
- •Selected Readings
- •Introduction
- •Why Write
- •Getting Started
- •Where and When to Write
- •Choosing the Journal
- •Instructions to Authors
- •Writing
- •Manuscript Writing Order
- •Figures and Tables
- •Methods
- •Results
- •Figure Legends
- •Introduction
- •Discussion
- •Acknowledgments
- •Abstract
- •Title
- •Authorship
- •Revising Before Submission
- •Responding to Reviewer Comments
- •References
- •Selected Readings
- •Introduction
- •Origins of the Term
- •Modern Definition and Primer
- •Transition from Mentee to Colleague
- •Mentoring Risks
- •Conclusion
- •References
- •Selected Readings
- •The Career Development Plan
- •Choosing the Mentor
- •Writing the Career Development Plan
- •The Candidate
- •Research Plan
- •Final Finishing Points About the Research Plan
- •Summary
- •References
- •Introduction
- •Decisions, Decisions!
- •Mission Impossible: Defining a Laboratory Mission or Vision
- •Project Planning
- •Saving Money
- •Seek Help
- •People
- •Who Should I Hire?
- •Advertising
- •References
- •Interviews
- •Conduct a Structured Interview
- •Probation Period
- •Trainees
- •Trainee Funding
- •Time Is on Your Mind
- •Research Techniques
- •Program Leadership
- •Summary
- •Selected Readings
- •Introduction
- •Direct Evidence
- •Indirect Evidence
- •Burnout
- •Prevention of and Recovery from Work–Life Imbalance
- •Action Plan for Finding Balance: Personal Level
- •Action Plan for Finding Balance: Professional Level
- •Conclusion
- •References
- •Introduction
- •Time Management Strategies
- •Planning and Prioritizing
- •Delegating and Saying “No”
- •Action Plans
- •Activity Logs
- •Scheduling Protected Time
- •Eliminating Distractions
- •Buffer Time
- •Goal Setting
- •Completing Large Tasks
- •Maximizing Efficiency
- •Get Organized
- •Multitasking
- •Think Positive
- •Summary
- •References
- •Selected Readings
- •Index
70 T.S. Riall
incidence of Crohn’s disease occurring most commonly in patients in their teens/twenties and patients in their sixties. For this type of distribution, the mean is not a useful or descriptive measure.
Measures of Spread
While the mean provides useful information, you will have more information if you have an idea of the spread (disper- sion or variation) of the observations around the mean. The range is the difference between the smallest and largest observation. It is common to give maximum and minimum values, which are more useful than the range. The range is used to emphasize extreme values.
Standard deviation (SD) is the most commonly used mea- sure of variation in medicine. It describes how observations cluster around the mean, and it is the basis for many statisti- cal tests used to compare means between groups. For each observation, the deviation from the mean is calculated and squared. The sum of the squared deviations for all observa- tions is divided by the number of observations minus one. This value is called the variance. The square root of the vari- ance is the SD. Regardless of the distribution of the data, at least 75% of observations lay between the mean plus 2 SDs and the mean minus 2 SDs. If a distribution is bell-shaped or normal, it has special characteristics: 67% of observations in a normal distribution lie between the mean ± 1 SD, 95% lie between the mean ± 2 SD,and 99.7% lie between the mean ± 3 SD. If the mean is smaller than two standard deviations, the data are probably skewed. Mean and SD are best used when the data are symmetric.
A percentile is the percentage of distribution that is at or below a particular number. Percentiles are commonly used to determine the “normal” ranges of laboratory values. Values lower than the 2.5 percentile and higher than the 97.5 percen- tile (greater or less than 2 SD from the mean) are considered outliers or abnormal values.The interquartile range is the dif- ference between the 25th and 75th percentile (or first and
Chapter 5. Analyzing Your Data |
71 |
third quartiles) and specifies the central 50% of observations. Percentiles and interquartile ranges are used when the median is used for ordinal or skewed data, or when the mean is used with the goal is to compare individual observations with a set of norms.
Comparison of Numeric Variables
Bivariate analysis is used to assess the relationship of a single independent variable (predictor) and a single dependent variable (outcome). Statistical tests for comparing means of continuous variables that are normally distributed include the student’s t-test for two independent groups and the paired t-test for paired samples. If the continuous variable is not normally distributed, nonparametric tests are used.These tests include the Wilcoxon rank sum test (also known as the Mann–Whitney U test) for two independent groups and the Wilcoxon signed-rank test for paired samples. Two groups of data are independent if the values in one group do not pro- vide any information about the values in the other group. Analysis of variance (ANOVA) is used to compare means among three or more normally distributed, independent groups. When comparing three or more groups, the P-value (corresponding to an F-test) indicates an overall significant difference and not differences between any two groups. To determine differences between any two groups, you need to do post-hoc comparison tests to perform multiple, pairwise comparisons including Tukey, Bonferroni, Newman–Keuls, and Fisher tests. The Kruskal–Wallis test is used to compare medians for three more independent groups in which the means are not normally distributed.
Comparison of Categorical Values
Categorical variables are expressed as proportions and can be demonstrated in 2 × 2 tables for two independent groups up to r x c tables for n independent groups. Chi-square tests