About

Karen Kafadar is a Commonwealth Professor in the Department of Statistics at the University of Virginia. She holds a Ph.D. in Statistics from Princeton University, an M.S. in Statistics from Stanford University, and a B.S. in Mathematics from Stanford University. Her research interests include robust methods and exploratory data analysis with applications to physical, chemical, engineering, and biological sciences, such as genomics, forensic science, randomized cancer screening trials, spatial data, and particle physics experiments.

Research topics

• Computer Science
• Sociology
• Social psychology
• Political Science
• Natural Language Processing
• Psychology
• Demography
• Medicine
• Criminology
• Cognitive psychology
• Econometrics
• Geography
• Statistics
• Management
• Economics
• Law
• Mathematics
• Environmental health

Selected publications

• Comment on “A Variance Decomposition Approach to Inconclusives in Forensic Black Box Studies” by Amanda Luby and Joseph Kadane

  Law Probability and Risk · 2026-01-01 · 1 citations

  article1st authorCorresponding
  PublisherDOI

• Robust Methods and Statistical Thinking for “Big Data” Science and Surveys

  Journal of Statistical Theory and Practice · 2026-02-08

  articleOpen access1st authorCorresponding

  Abstract Data science and machine learning algorithms are sometimes viewed as the only tools that are needed to analyze large datasets. Yet concepts from classical statistics remain critical in such settings. Massive data are rarely independent, outlier-free, or homogeneous: clusters, subdomains of observations, multiplicity of tests, and hidden trends are common and require statistical thinking, robust methods, and insightful displays. Sampling methodology, along with survey design and analysis, are essential in our current statistical framework for ensuring valid inferences with quantifiable uncertainties. This paper discusses some datasets where statistical analysis uncovered subtle biases and discrepancies that would have been hidden in these seemingly trustworthy, data-rich sources. Until a new statistical framework is developed to generate valid inferences on non-randomized, highly dependent clustered data, these examples demonstrate that statistical thinking, statistical methods, and informative displays remain critical for ensuring valid analyses and communication of justified conclusions from “Big Data.”

  PublisherOA PDFDOI

• Enhancing foundational validity of forensic findings in nonlethal medico-legal strangulation examinations

  Journal of Forensic and Legal Medicine · 2025-01-01 · 1 citations

  article
  PublisherDOI

• Persistence of the verbal overshadowing and weapon-focus effects on lineup identification performance.

  Journal of Applied Research in Memory and Cognition · 2024-11-18 · 2 citations

  article

  Peer reviewed

  PublisherDOI

• Challenges in Modeling, Interpreting, and Drawing Conclusions from Images as Forensic Evidence

  Statistics and data science in imaging. · 2024-01-01 · 2 citations

  articleOpen access1st authorCorresponding
  PublisherDOI

• Sensitizing jurors to eyewitness confidence using “reason-based” judicial instructions.

  Journal of Applied Research in Memory and Cognition · 2022-06-23 · 13 citations

  article
  PublisherDOI

• The q–q Boxplot

  Iowa State University Digital Repository (Iowa State University) · 2021-07-19

  articleSenior author

  Boxplots have become an extremely popular display of distribution summaries for collections of data, especially when we need to visualize summaries for several collections simultaneously. The whiskers in the boxplot show only the extent of the tails for most of the data (with outside values denoted separately); more detailed information about the shape of the tails, such as skewness and \weight" relative to a standard reference distribution, is much better displayed via quantile-quantile (q-q) plots. We incorporate the q-q plot's tail information into the traditional boxplot by replacing the boxplot's whiskers with the tails from a q-q plot, and display these tails with con dence bands for the tails that would be expected from the tails of the reference distribution. We describe the construction of the "q-q boxplot" and demonstrate its advantages over earlier proposed boxplot modi cations on data from economics and neuroscience, which illustrate q-q boxplots' effectiveness in showing important tail behavior especially for large datasets. The package qqboxplot (an extension to the ggplot2 package (Wickham, 2016)) is available for the R (R Core Team, 2020) programming language.

  Publisher

• The ASA president’s task force statement on statistical significance and replicability

  The Annals of Applied Statistics · 2021 · 40 citations

  Senior authorCorresponding
  • Computer Science
  • Natural Language Processing
  • Political Science

  Over the past decade, the sciences have experienced elevated concerns about the replicability of study results. An important aspect of replicability is the use of statistical methods for framing conclusions. In 2019 the President of the American Statistical Association (ASA) established a task force to address concerns that a 2019 editorial in The American Statisti cian (an ASA journal) might be mistakenly interpreted as official ASA policy. (The 2019 editorial recommended eliminating the use of “p < 0.05” and “statistically significant” in statistical analysis.) This document is the statement of the task force, and the ASA invited us to publicize it. Its purpose is two-fold: to clarify that the use of P -values and significance testing, properly applied and interpreted, are important tools that should not be abandoned, and to briefly set out some principles of sound statistical inference that may be useful to the scientific community.

  PublisherOA PDFDOI

• Eyewitness identification speed: Slow identifications from highly confident eyewitnesses hurt perceptions of their testimony.

  Journal of Applied Research in Memory and Cognition · 2021-01-18 · 8 citations

  article

  How persuasive is eyewitness confidence? Are highly confident eyewitnesses so persuasive that their testimony overshadows other countervailing evidence? To answer these questions, participants evaluated a highly confident eyewitness’s lineup identification. Participants learned that an eyewitness either quickly identified the suspect (e.g., “I’m sure it’s him. I identified him instantly.”), slowly identified the suspect (e.g., “I’m sure it’s him. I identified him after a while.”) or they learned nothing about the eyewitness’s identification time (e.g., “I’m sure it’s him.”). Highly confident eyewitnesses who make a relatively slow identification are perceived as less accurate and suspects are regarded as less likely to be guilty as compared to when eyewitnesses make a fast identification or even when no information is provided about identification speed. Identification speed appears to be one of the few variables that can cause people to regard with skepticism the testimony of highly confident eyewitnesses.

  PublisherDOI

• Some Basic Statistical Methods for Chromatographic Data

  Advances in chromatography · 2021-06-22 · 2 citations

  book-chapter1st authorCorresponding

  The application of statistical methods has been increasing over the past decades, particularly in the fields of engineering and physical and chemical sciences. Many statistical concepts were formulated originally for biology or agricultural experiments, but their importance has reached all branches of science, since measurement methods for determining physical, biological, or behavioral relationships always involve some variability. Statistics is the science of collecting, analyzing, and interpreting numerical data. An important goal of statistics is to assess the magnitude of fluctuations in the data from various sources, thereby gaining understanding of a measurement process. Statistical inference, properly applied, aids in this generalization. Furthermore, a well-planned, statistically designed experiment will maximize the amount of information per measurement, thereby reducing the cost of unnecessary work.

  PublisherDOI

Frequent coauthors

• Bernice A. Pescosolido

  Indiana University Bloomington

  79 shared

• Byungkyu Lee

  78 shared

• Brandon L. Garrett

  Duke University

  13 shared

• Joanne Yaffe

  11 shared

• Philip C. Prorok

  National Institutes of Health

  11 shared

• Chad S. Dodson

  University of Virginia

  10 shared

• Kitty Corbett

  Simon Fraser University

  9 shared

• Shale Wong

  8 shared