Loading
Expert Witness Guides
12 min read
Areas of Expertise :
Statistics, Data Science, Biostatistics, Economics, Mathematics
Published Date:
Last Modified Date:
Statistical expert testimony must do more than present complex calculations; courts evaluate whether the data, methodology, assumptions, and case-specific application are sufficiently reliable under Federal Rule of Evidence 702. This article explains how courts assess statistical experts, including issues involving sample size, selection bias, confounding variables, regression models, statistical significance, transparency, and reproducibility.
Statistical evidence plays an important role in modern litigation, particularly when courts must evaluate complex economic, scientific, technical, or other specialized issues. Experts use statistical analyses in employment disputes, economic damages claims, antitrust cases, environmental litigation, public health matters, consumer class actions, and other data-intensive controversies.
Presenting a statistical analysis, however, does not automatically make an expert’s opinion admissible. Under Federal Rule of Evidence 702, the court evaluates whether the testimony will help the trier of fact, is based on sufficient facts or data, is the product of reliable principles and methods, and reflects a reliable application of those methods to the facts of the case.
Rule 702 principally governs expert testimony and opinions. A written expert report may explain those opinions and their supporting analysis, but the report itself is not automatically admissible as an exhibit merely because the testimony satisfies Rule 702. Other evidentiary rules may affect whether the written report can be introduced at trial.
This article focuses primarily on the federal framework. Rule 702, together with the Supreme Court’s decisions in Daubert v. Merrell Dow Pharmaceuticals, Inc., General Electric Co. v. Joiner, and Kumho Tire Co. v. Carmichael, provides the foundation for evaluating statistical expert evidence in federal court.
State courts may apply Rule 702, a state-law equivalent, the Daubert framework, the Frye standard, or another jurisdiction-specific test. Attorneys should always confirm the governing evidentiary standard in the relevant jurisdiction.
Although this article discusses scientific reliability, Rule 702 applies broadly to scientific, technical, and other specialized knowledge. The specific reliability inquiry depends on the expert’s discipline, methodology, data, and opinions.
Not every case involving numerical evidence requires a statistician. When an expert’s opinions depend substantially on statistical analysis, however, the proponent must establish that the opinions satisfy the governing admissibility requirements.
Statistical evidence commonly appears in cases involving:
Courts evaluate the expert’s qualifications, the sufficiency of the underlying facts or data, the reliability of the methodology, whether the methodology was reliably applied, and whether the testimony will help the trier of fact.
Courts do not decide at the admissibility stage which qualified expert is ultimately correct. When qualified experts use reliable methods but reasonably reach different conclusions, their disagreement may affect the weight of the evidence rather than admissibility. Each opinion must still independently satisfy Rule 702.
Regression Analysis
Quantitative analysis
Data Analysis
Econometrics
Analytics
Predictive Analytics
Statistical evidence plays an important role in modern litigation, particularly when courts must evaluate complex economic, scientific, technical, or other specialized issues. Experts use statistical analyses in employment disputes, economic damages claims, antitrust cases, environmental litigation, public health matters, consumer class actions, and other data-intensive controversies.
Presenting a statistical analysis, however, does not automatically make an expert’s opinion admissible. Under Federal Rule of Evidence 702, the court evaluates whether the testimony will help the trier of fact, is based on sufficient facts or data, is the product of reliable principles and methods, and reflects a reliable application of those methods to the facts of the case.
Rule 702 principally governs expert testimony and opinions. A written expert report may explain those opinions and their supporting analysis, but the report itself is not automatically admissible as an exhibit merely because the testimony satisfies Rule 702. Other evidentiary rules may affect whether the written report can be introduced at trial.
This article focuses primarily on the federal framework. Rule 702, together with the Supreme Court’s decisions in Daubert v. Merrell Dow Pharmaceuticals, Inc., General Electric Co. v. Joiner, and Kumho Tire Co. v. Carmichael, provides the foundation for evaluating statistical expert evidence in federal court.
State courts may apply Rule 702, a state-law equivalent, the Daubert framework, the Frye standard, or another jurisdiction-specific test. Attorneys should always confirm the governing evidentiary standard in the relevant jurisdiction.
Although this article discusses scientific reliability, Rule 702 applies broadly to scientific, technical, and other specialized knowledge. The specific reliability inquiry depends on the expert’s discipline, methodology, data, and opinions.
Not every case involving numerical evidence requires a statistician. When an expert’s opinions depend substantially on statistical analysis, however, the proponent must establish that the opinions satisfy the governing admissibility requirements.
Statistical evidence commonly appears in cases involving:
Courts evaluate the expert’s qualifications, the sufficiency of the underlying facts or data, the reliability of the methodology, whether the methodology was reliably applied, and whether the testimony will help the trier of fact.
Courts do not decide at the admissibility stage which qualified expert is ultimately correct. When qualified experts use reliable methods but reasonably reach different conclusions, their disagreement may affect the weight of the evidence rather than admissibility. Each opinion must still independently satisfy Rule 702.
Federal Rule of Evidence 702 governs the admissibility of expert testimony in federal courts. The proponent must establish that:
These requirements operate together. Satisfying one requirement does not excuse failure to satisfy another.
The amendment to Rule 702 that took effect on December 1, 2023, clarified the existing Rule 104(a) requirement that the proponent establish admissibility by a preponderance of the evidence. It also emphasized that each expert opinion must remain within what can reliably be concluded from the expert’s factual basis and methodology.
The Supreme Court’s decisions in Daubert, Joiner, and Kumho Tire continue to guide federal courts.
In Daubert, the Court held that trial judges must act as gatekeepers by evaluating the reliability and relevance of expert testimony before admitting it.
In Joiner, the Court confirmed that appellate courts generally review expert-admissibility decisions for abuse of discretion. It also recognized that a court may exclude an opinion when there is too great an analytical gap between the underlying data and the expert’s conclusion.
In Kumho Tire, the Court explained that the gatekeeping obligation applies to scientific, technical, and other specialized knowledge. Statistical experts are therefore subject to Rule 702, although the specific reliability considerations may vary according to the discipline, method, and opinion at issue.
Reliability focuses on how the expert reached the opinion, not simply on whether one party agrees with the conclusion.
A reliable statistical analysis generally:
Qualified experts may interpret the same dataset differently while using reliable methods. The court considers whether each opinion is supported by sufficient facts or data and reflects a reliable application of appropriate methods.
Relevant expertise may include statistics, biostatistics, econometrics, epidemiology, economics, mathematics, data science, or another quantitative discipline related to the opinions offered.
Academic credentials alone do not establish qualification. Courts may also consider education, training, professional experience, publications, research, teaching, certifications, and practical work. The expert must be qualified for the particular opinions offered, not necessarily for every discipline implicated by the case.
Even a highly qualified expert must demonstrate that the methodology and its application satisfy Rule 702.
The Daubert decision identified several flexible considerations that may assist the court, including whether the method:
No single factor controls admissibility. The appropriate inquiry depends on the method, discipline, and issues presented.
Peer review does not automatically establish that a case-specific opinion is reliable. Courts may also examine the underlying data, assumptions, analytical decisions, and application of the method.
An accepted statistical method may still produce an unreliable opinion if it is applied incorrectly.
Courts may consider whether the expert:
For example, a regression analysis may become less persuasive, and in some circumstances unreliable for the opinion offered, if it excludes an important explanatory variable without adequate justification.
The question is not whether every possible variable was included. The court instead considers whether omitted variables materially affect the model’s ability to answer the relevant legal question.
Reliable statistical methods cannot, by themselves, cure material defects in the underlying data. Some data limitations can be addressed through appropriate statistical procedures, but the expert must explain those limitations and show that the available information sufficiently supports the opinions offered.
Courts may consider:
Not every data problem makes expert testimony inadmissible. Minor deficiencies may affect the weight of the evidence. Material flaws that undermine the foundation of an opinion, however, may support exclusion.
For attorneys, reviewing statistical evidence requires evaluating both the analytical method and the quality of the underlying data.
A larger dataset does not automatically produce a more reliable conclusion. Sample size is only one part of the analysis, and there is no universal minimum sample size that guarantees admissibility.
Relevant considerations include:
A properly selected larger sample may improve precision and the ability to detect meaningful effects. A large but biased or poorly designed sample may still produce unreliable results.
A well-prepared analysis should explain how sample size, study design, and available data affect statistical uncertainty, precision, and the permissible scope of the conclusions.
Selection bias can arise when the process used to select observations, participants, records, treatment groups, or complete cases creates systematic differences that distort the relationship being studied.
One common form occurs when the analyzed sample does not adequately represent the population relevant to the dispute.
Examples include:
Courts may examine how information was collected, selected, verified, and analyzed when determining whether an opinion satisfies Rule 702.
In causal analysis, a confounding variable is generally a factor associated with both the explanatory variable or exposure of interest and the outcome. Failing to account for it may distort the estimated relationship between them.
For example, wage differences in an employment case may be associated with education, experience, geographic location, job responsibilities, specialized training, or other legitimate factors.
Experts are not required to include every variable associated with an outcome. The important question is whether omitted variables materially impair the model’s ability to answer the legal question.
Statistical significance is frequently misunderstood in litigation.
It is commonly evaluated using a p-value, which indicates how incompatible the observed data are with a specified statistical model, often a null hypothesis, under the assumptions used in the analysis.
A statistically significant result does not establish that a hypothesis is true. It also does not independently prove causation, liability, or damages.
Courts and experts should distinguish among:
These concepts answer different questions. A statistically significant result may have little practical importance. Statistical or practical significance alone does not establish legal liability.
The legal importance of statistical evidence depends on the substantive claim, the element at issue, the purpose for which the evidence is offered, and the remaining evidentiary record.
Regression analysis estimates relationships among variables while accounting for other measurable factors.
Courts frequently encounter regression analysis in employment cases, economic damages matters, antitrust disputes, consumer class actions, and other quantitative litigation.
Courts may consider whether:
Technical complexity alone does not establish reliability. The expert must explain why the method is appropriate and how it supports the opinions offered.
Rule 702 also applies when expert opinions rely on Bayesian analysis, time-series analysis, survival analysis, machine learning, predictive modeling, or other quantitative techniques. The reliability inquiry should be adapted to the method, data, validation procedures, intended use, and opinion offered.
Transparency allows courts and opposing experts to evaluate whether an opinion satisfies Rule 702.
Well-prepared analyses typically:
Reproducing the same software output does not, by itself, establish reliability. The expert must also explain the data, assumptions, model specifications, coding decisions, and interpretation.
Whether source code, proprietary databases, confidential records, or protected materials must be disclosed may depend on discovery rules, protective orders, licensing restrictions, and case-specific court orders.
Common Rule 702 challenges allege that:
Courts should not treat every dispute over an expert’s factual basis or application as automatically going only to weight.
The judge must first determine under Rule 104(a) whether the proponent has shown that Rule 702’s requirements are more likely than not satisfied. Issues remaining after admissibility is established may then be addressed through cross-examination, competing expert testimony, and other evidence.
Statistical experts do more than perform calculations. They help judges and juries understand quantitative evidence and explain the strengths and limitations of their analyses.
An effective statistical expert can:
Experts may strengthen their credibility by acknowledging meaningful limitations rather than overstating the certainty of their conclusions.
Statistical evidence plays an important role in modern litigation, but sophisticated calculations alone do not satisfy Rule 702.
When admissibility is disputed, the court must determine whether the challenged opinions are based on sufficient facts or data, rely on reliable principles and methods, reflect a reliable application of those methods, and will help the trier of fact.
For attorneys, these principles provide a framework for expert selection, expert discovery, and Rule 702 motion practice. They can help counsel identify methodological weaknesses in an opposing analysis and determine whether a retained expert can adequately explain and defend the opinions offered.
For statistical experts, reliable methodology is only one part of persuasive testimony. A clear explanation of the analysis, assumptions, limitations, and case-specific application can help demonstrate that the opinions satisfy Rule 702.
Ultimately, reliable statistical expert evidence does more than produce numerical results. It presents a supported, transparent, and logically reasoned analysis that helps the court evaluate complex facts under the governing evidentiary standard.
Can a statistically significant result prove liability?
No. Statistical significance does not, by itself, establish causation, liability, or damages. Its legal effect depends on the substantive claim, the element at issue, the methodology used, and the remaining evidence.
Are peer-reviewed statistical methods always required?
No. Peer review is one flexible Daubert consideration, not an absolute requirement. Courts evaluate reliability based on the method, discipline, intended use, and circumstances of the case.
Can two qualified statisticians reach different conclusions?
Yes. Qualified experts may reasonably interpret the same data differently while using reliable methods. Each opinion must independently satisfy Rule 702.
Can a court exclude a statistical expert’s opinions even if the expert is highly qualified?
Yes. Qualifications alone do not establish admissibility. The proponent must also show that the opinions are based on sufficient facts or data, result from reliable principles and methods, reflect a reliable application of those methods, and will help the trier of fact.
Federal Rule of Evidence 702 governs the admissibility of expert testimony in federal courts. The proponent must establish that:
These requirements operate together. Satisfying one requirement does not excuse failure to satisfy another.
The amendment to Rule 702 that took effect on December 1, 2023, clarified the existing Rule 104(a) requirement that the proponent establish admissibility by a preponderance of the evidence. It also emphasized that each expert opinion must remain within what can reliably be concluded from the expert’s factual basis and methodology.
The Supreme Court’s decisions in Daubert, Joiner, and Kumho Tire continue to guide federal courts.
In Daubert, the Court held that trial judges must act as gatekeepers by evaluating the reliability and relevance of expert testimony before admitting it.
In Joiner, the Court confirmed that appellate courts generally review expert-admissibility decisions for abuse of discretion. It also recognized that a court may exclude an opinion when there is too great an analytical gap between the underlying data and the expert’s conclusion.
In Kumho Tire, the Court explained that the gatekeeping obligation applies to scientific, technical, and other specialized knowledge. Statistical experts are therefore subject to Rule 702, although the specific reliability considerations may vary according to the discipline, method, and opinion at issue.
Reliability focuses on how the expert reached the opinion, not simply on whether one party agrees with the conclusion.
A reliable statistical analysis generally:
Qualified experts may interpret the same dataset differently while using reliable methods. The court considers whether each opinion is supported by sufficient facts or data and reflects a reliable application of appropriate methods.
Relevant expertise may include statistics, biostatistics, econometrics, epidemiology, economics, mathematics, data science, or another quantitative discipline related to the opinions offered.
Academic credentials alone do not establish qualification. Courts may also consider education, training, professional experience, publications, research, teaching, certifications, and practical work. The expert must be qualified for the particular opinions offered, not necessarily for every discipline implicated by the case.
Even a highly qualified expert must demonstrate that the methodology and its application satisfy Rule 702.
The Daubert decision identified several flexible considerations that may assist the court, including whether the method:
No single factor controls admissibility. The appropriate inquiry depends on the method, discipline, and issues presented.
Peer review does not automatically establish that a case-specific opinion is reliable. Courts may also examine the underlying data, assumptions, analytical decisions, and application of the method.
An accepted statistical method may still produce an unreliable opinion if it is applied incorrectly.
Courts may consider whether the expert:
For example, a regression analysis may become less persuasive, and in some circumstances unreliable for the opinion offered, if it excludes an important explanatory variable without adequate justification.
The question is not whether every possible variable was included. The court instead considers whether omitted variables materially affect the model’s ability to answer the relevant legal question.
Reliable statistical methods cannot, by themselves, cure material defects in the underlying data. Some data limitations can be addressed through appropriate statistical procedures, but the expert must explain those limitations and show that the available information sufficiently supports the opinions offered.
Courts may consider:
Not every data problem makes expert testimony inadmissible. Minor deficiencies may affect the weight of the evidence. Material flaws that undermine the foundation of an opinion, however, may support exclusion.
For attorneys, reviewing statistical evidence requires evaluating both the analytical method and the quality of the underlying data.
A larger dataset does not automatically produce a more reliable conclusion. Sample size is only one part of the analysis, and there is no universal minimum sample size that guarantees admissibility.
Relevant considerations include:
A properly selected larger sample may improve precision and the ability to detect meaningful effects. A large but biased or poorly designed sample may still produce unreliable results.
A well-prepared analysis should explain how sample size, study design, and available data affect statistical uncertainty, precision, and the permissible scope of the conclusions.
Selection bias can arise when the process used to select observations, participants, records, treatment groups, or complete cases creates systematic differences that distort the relationship being studied.
One common form occurs when the analyzed sample does not adequately represent the population relevant to the dispute.
Examples include:
Courts may examine how information was collected, selected, verified, and analyzed when determining whether an opinion satisfies Rule 702.
In causal analysis, a confounding variable is generally a factor associated with both the explanatory variable or exposure of interest and the outcome. Failing to account for it may distort the estimated relationship between them.
For example, wage differences in an employment case may be associated with education, experience, geographic location, job responsibilities, specialized training, or other legitimate factors.
Experts are not required to include every variable associated with an outcome. The important question is whether omitted variables materially impair the model’s ability to answer the legal question.
Statistical significance is frequently misunderstood in litigation.
It is commonly evaluated using a p-value, which indicates how incompatible the observed data are with a specified statistical model, often a null hypothesis, under the assumptions used in the analysis.
A statistically significant result does not establish that a hypothesis is true. It also does not independently prove causation, liability, or damages.
Courts and experts should distinguish among:
These concepts answer different questions. A statistically significant result may have little practical importance. Statistical or practical significance alone does not establish legal liability.
The legal importance of statistical evidence depends on the substantive claim, the element at issue, the purpose for which the evidence is offered, and the remaining evidentiary record.
Regression analysis estimates relationships among variables while accounting for other measurable factors.
Courts frequently encounter regression analysis in employment cases, economic damages matters, antitrust disputes, consumer class actions, and other quantitative litigation.
Courts may consider whether:
Technical complexity alone does not establish reliability. The expert must explain why the method is appropriate and how it supports the opinions offered.
Rule 702 also applies when expert opinions rely on Bayesian analysis, time-series analysis, survival analysis, machine learning, predictive modeling, or other quantitative techniques. The reliability inquiry should be adapted to the method, data, validation procedures, intended use, and opinion offered.
Transparency allows courts and opposing experts to evaluate whether an opinion satisfies Rule 702.
Well-prepared analyses typically:
Reproducing the same software output does not, by itself, establish reliability. The expert must also explain the data, assumptions, model specifications, coding decisions, and interpretation.
Whether source code, proprietary databases, confidential records, or protected materials must be disclosed may depend on discovery rules, protective orders, licensing restrictions, and case-specific court orders.
Common Rule 702 challenges allege that:
Courts should not treat every dispute over an expert’s factual basis or application as automatically going only to weight.
The judge must first determine under Rule 104(a) whether the proponent has shown that Rule 702’s requirements are more likely than not satisfied. Issues remaining after admissibility is established may then be addressed through cross-examination, competing expert testimony, and other evidence.
Statistical experts do more than perform calculations. They help judges and juries understand quantitative evidence and explain the strengths and limitations of their analyses.
An effective statistical expert can:
Experts may strengthen their credibility by acknowledging meaningful limitations rather than overstating the certainty of their conclusions.
Statistical evidence plays an important role in modern litigation, but sophisticated calculations alone do not satisfy Rule 702.
When admissibility is disputed, the court must determine whether the challenged opinions are based on sufficient facts or data, rely on reliable principles and methods, reflect a reliable application of those methods, and will help the trier of fact.
For attorneys, these principles provide a framework for expert selection, expert discovery, and Rule 702 motion practice. They can help counsel identify methodological weaknesses in an opposing analysis and determine whether a retained expert can adequately explain and defend the opinions offered.
For statistical experts, reliable methodology is only one part of persuasive testimony. A clear explanation of the analysis, assumptions, limitations, and case-specific application can help demonstrate that the opinions satisfy Rule 702.
Ultimately, reliable statistical expert evidence does more than produce numerical results. It presents a supported, transparent, and logically reasoned analysis that helps the court evaluate complex facts under the governing evidentiary standard.
Can a statistically significant result prove liability?
No. Statistical significance does not, by itself, establish causation, liability, or damages. Its legal effect depends on the substantive claim, the element at issue, the methodology used, and the remaining evidence.
Are peer-reviewed statistical methods always required?
No. Peer review is one flexible Daubert consideration, not an absolute requirement. Courts evaluate reliability based on the method, discipline, intended use, and circumstances of the case.
Can two qualified statisticians reach different conclusions?
Yes. Qualified experts may reasonably interpret the same data differently while using reliable methods. Each opinion must independently satisfy Rule 702.
Can a court exclude a statistical expert’s opinions even if the expert is highly qualified?
Yes. Qualifications alone do not establish admissibility. The proponent must also show that the opinions are based on sufficient facts or data, result from reliable principles and methods, reflect a reliable application of those methods, and will help the trier of fact.
Regression Analysis
Quantitative analysis
Data Analysis
Econometrics
Analytics
Predictive Analytics
Expert Witness Guides
Expert Witness Guides