About

Philip B. Stark is a Distinguished Professor in the Department of Statistics at the University of California, Berkeley. His research centers on inference problems primarily in physical and social sciences, with a particular interest in confidence procedures tailored for specific goals and in quantifying the uncertainty in inferences that rely on numerical models of complex systems. His work encompasses a broad range of topics including uncertainty quantification and inference, inverse problems, nonparametrics, risk assessment, elections, geophysics, astrophysics, cosmology, litigation, and health. Professor Stark has conducted research on the internal structure of the Sun and Earth, climate modeling, clinical trials, earthquake prediction, the Big Bang, the geomagnetic field, election integrity, gender bias in academia, geriatric hearing loss, the U.S. census, the effectiveness of Internet content filters, endangered species, spectrum estimation, urban foraging, and information retrieval. He is also interested in nutrition, food equity, and sustainability, studying whether foraging wild foods in urban environments could contribute to nutrition, especially in food deserts. His expertise extends to numerical optimization, and he has published software related to his research. In addition to his research, Stark has consulted in various legal and policy contexts, including product liability litigation, truth in advertising, election security, contested elections, trade secret litigation, employment discrimination, environmental litigation, patent litigation, sampling in litigation, and the U.S. census. He created SticiGui, an online introductory statistics resource that includes interactive data analysis, demonstrations, machine-graded online assignments, and exercises, which was the basis of the first online course taught at UC Berkeley. Alongside Ani Adhikari, he co-taught a popular introductory statistics MOOC in 2013, which enrolled over 52,600 students, with more than 10,600 completing the course and nearly 8,200 receiving certificates.

Research topics

• Political Science
• Computer Security
• Social Science
• Computer Science
• Sociology
• Law
• Engineering
• Law and economics
• Public relations
• Medicine
• Social psychology
• Psychology

Selected publications

• Fast Conservative Monte Carlo Confidence Sets

  Journal of Computational and Graphical Statistics · 2025-07-25

  articleSenior author
  PublisherDOI

• Generalization and Power of Kocay's Lemma in Graph Reconstruction

  ArXiv.org · 2025-08-29

  preprintOpen access1st authorCorresponding

  This paper generalizes Kocay's lemma, with particular applications to graph reconstruction, as well as discussing and proving aspects around the power of these generalizations and Kocay's original lemma, with a result on the reconstruction of the multiplicity of a tree T as a subgraph of G.

  PublisherOA PDFDOI

• Doing More with Less: Mismatch-Based Risk-Limiting Audits

  Lecture notes in computer science · 2025-11-05

  book-chapter
  PublisherDOI

• Dice, but Don’t Slice: Optimizing the Efficiency of ONEAudit

  Lecture notes in computer science · 2025-09-10

  book-chapterOpen accessSenior author

  Abstract ONEAudit provides more efficient risk-limiting audits than other extant methods when the voting system cannot report a cast-vote record linked to each cast card. It obviates the need for re-scanning; it is simpler and more efficient than ‘hybrid’ audits; and it is far more efficient than batch-level comparison audits. There may be room to improve the efficiency of ONEAudit further by tuning the statistical tests it uses and by using stratified sampling. We show that tuning the tests by optimizing for the reported batch-level tallies or integrating over a distribution reduces expected workloads by 70–85% compared to the current ONEAudit implementation across a range of simulated elections. The improved tests reduce the expected workload to audit the 2024 Mayoral race in San Francisco, California, by half—from about 200 cards to about 100 cards. In contrast, stratified sampling does not help: it increases workloads by about 25% on average.

  PublisherOA PDFDOI

• Doing More With Less: Mismatch-Based Risk-Limiting Audits

  ArXiv.org · 2025-03-20

  preprintOpen access

  One approach to risk-limiting audits (RLAs) compares randomly selected cast vote records (CVRs) to votes read by human auditors from the corresponding ballot cards. Historically, such methods reduce audit sample sizes by considering how each sampled CVR differs from the corresponding true vote, not merely whether they differ. Here we investigate the latter approach, auditing by testing whether the total number of mismatches in the full set of CVRs exceeds the minimum number of CVR errors required for the reported outcome to be wrong (the "CVR margin"). This strategy makes it possible to audit more social choice functions and simplifies RLAs conceptually, which makes it easier to explain than some other RLA approaches. The cost is larger sample sizes. "Mismatch-based RLAs" only require a lower bound on the CVR margin, which for some social choice functions is easier to calculate than the effect of particular errors. When the population rate of mismatches is low and the lower bound on the CVR margin is close to the true CVR margin, the increase in sample size is small. However, the increase may be very large when errors include errors that, if corrected, would widen the CVR margin rather than narrow it; errors affect the margin between candidates other than the reported winner with the fewest votes and the reported loser with the most votes; or errors that affect different margins.

  PublisherOA PDFDOI

• The Use of Airway Pressure Release Ventilation in a Patient With Severe Respiratory Alkalosis Following Intubation

  American Journal of Respiratory and Critical Care Medicine · 2025-05-01

  article

  Abstract While protocols for management of respiratory acidosis in the ventilated patient have been well described, management in the setting of significant respiratory alkalosis in the vent dependent patient are limited. Typical management includes use of acetazolamide, although its use has not been shown to reduce mortality in ventilated patients. Other methods such as use of benzodiazepines and opiates to depress respiration have been described, but poorly studied. This patient is a 76-year-old man with a significant cardiac disease and chronic obstructive pulmonary disease requiring 4L of oxygen at baseline, who presented to the emergency department with increased work of breathing, lethargy, and altered mental status. The patient had been managed at home for a heart failure exacerbation and had developed a significant contraction alkalosis. On presentation to the ED the patient was hemodynamically stable, but venous blood gas (VBG) showed hypercarbia of 109.2 mmHg with a relatively normal pH (7.303), and an elevated bicarbonate (54mmol/L). The oxygen saturation was 100% on 6L nasal cannula. Imaging upon presentation showed pulmonary edema with mild bilateral pulmonary effusion. Noninvasive ventilation with BiPAP produced minimal change and due to concern for worsening respiratory decompensation, the patient was intubated in the ED. Following intubation, recheck of patient's VBG showed marked alkalosis (pH 7.68, pCO2 43). Upon arrival to the ICU, high levels of analgesia, sedation and acetazolamide were trialed in an attempt to decrease the pH and increase the pCO2. The patient's vent was placed in airway pressure release ventilation (APRV) mode in an attempt to decrease ventilation while preserving oxygenation in the setting of refractory alkalosis. The patient did see resolution of his alkalosis with the change in vent setting. On HD 6 Patient's pH stabilized and sedation was weaned. Afterwards patient was successfully extubated to HFNC. The patient passed after 14 days due to continued complications. The use of APRV mode has been used in the setting of ARDS to help with oxygenation with known reduced ventilation. It has not been studied specifically in the context of isolated respiratory alkalosis. Here we present a novel method to control respiratory alkalosis with high Thigh to minimize releases, and a low Phigh, which results in lower minute ventilation. Using invasive ventilation modes that maintain oxygenation but reduce ventilation may be worth considering in the future in patients with difficult to control respiratory alkalosis.

  PublisherDOI

• Estimating soil carbon sequestration potential and approximating optimal management policies

  arXiv (Cornell University) · 2024-09-26

  preprintOpen access

  The impact of a management intervention on the soil organic carbon (SOC) stored in a given volume of soil is moderated by features that determine that soil's sequestration potential under that intervention. To maximize total SOC sequestration cost efficiently, interventions should be targeted to soils with the highest responses and lowest intervention costs. We present a framework for estimating SOC sequestration potentials and approximating efficient management policies. We review relevant sources of measurement uncertainty and formalize policy choice using potential outcomes. An optimal sequestration policy can be approximated by modeling SOC measurements as functions of covariates within each treatment group, using the fitted models to estimate SOC sequestration potential for each plot, and finding the policy that maximizes the average of those estimates. The modeling can use linear regression or other algorithms to learn relationships between features and SOC sequestration potential. We demonstrate this method using data from a study of compost amendments applied to California rangelands. We find that the plots exhibit treatment effects moderated by baseline SOC -- so targeting amendments to plots with lower baseline SOC would increase overall SOC sequestration rates. We evaluate these methods further in simulated field experiments. Refined policy estimates sequestered more SOC than uniform application of the single management policy estimated to have the largest average treatment effect, especially when SOC sequestration potential could be predicted from observed features. We conclude by discussing baseline SOC moderation, observational studies, inference, cost models, and broader policy uncertainties.

  PublisherOA PDFDOI

• Improving the Computational Efficiency of Adaptive Audits of IRV Elections

  Lecture notes in computer science · 2024-09-22 · 1 citations

  book-chapterOpen access

  Abstract AWAIRE is one of two extant methods for conducting risk-limiting audits of instant-runoff voting (IRV) elections. In principle AWAIRE can audit IRV contests with any number of candidates, but the original implementation incurred memory and computation costs that grew superexponentially with the number of candidates. This paper improves the algorithmic implementation of AWAIRE in three ways that make it practical to audit IRV contests with 55 candidates, compared to the previous 6 candidates. First, rather than trying from the start to rule out all candidate elimination orders that produce a different winner, the algorithm starts by considering only the final round, testing statistically whether each candidate could have won that round. For those candidates who cannot be ruled out at that stage, it expands to consider earlier and earlier rounds until either it provides strong evidence that the reported winner really won or a full hand count is conducted, revealing who really won. Second, it tests a richer collection of conditions, some of which can rule out many elimination orders at once. Third, it exploits relationships among those conditions, allowing it to abandon testing those that are unlikely to help. We provide real-world examples with up to 36 candidates and synthetic examples with up to 55 candidates, showing how audit sample size depends on the margins and on the tuning parameters. An open-source Python implementation is publicly available.

  PublisherOA PDFDOI

• Efficient Weighting Schemes for Auditing Instant-Runoff Voting Elections

  Lecture notes in computer science · 2024-11-29 · 1 citations

  preprintOpen access
  PublisherOA PDFDOI

• Sequential stratified inference for the mean

  ArXiv.org · 2024-09-10

  preprintOpen accessSenior author

  We develop conservative tests for the mean of a bounded population under stratified sampling and apply them to risk-limiting post-election audits. The tests are ``anytime valid'' under sequential sampling, allowing optional stopping in each stratum. Our core method expresses a global hypothesis about the population mean as a union of intersection hypotheses describing within-stratum means. It tests each intersection hypothesis using independent test supermartingales (TSMs) combined across strata by multiplication. A $P$-value for each intersection hypothesis is the reciprocal of that test statistic, and the largest $P$-value in the union is a $P$-value for the global hypothesis. This approach has two primary moving parts: the rule selecting which stratum to draw from next given the sample so far, and the form of the TSM within each stratum. These rules may vary over intersection hypotheses. We construct the test with the smallest expected stopping time and present a few strategies for approximating that optimum. In instances that arise in auditing and other applications, its expected sample size is substantially smaller than that of previous methods.

  PublisherOA PDFDOI

Frequent coauthors

• Vanessa Teague

  Australian National University

  37 shared

• Ronald L. Rivest

  25 shared

• Josh Benaloh

  Microsoft (United States)

  21 shared

• Damjan Vukcevic

  19 shared

• Jasjeet S. Sekhon

  17 shared

• Dan S. Wallach

  Defense Advanced Research Projects Agency

  16 shared

• Kellie Ottoboni

  University of California, Berkeley

  16 shared

• Peter J. Stuckey

  16 shared