Publications and Preprints

@misc{bhaduri2025conditioning,
  title={Conditioning on posterior samples for flexible frequentist goodness-of-fit testing},
  author={Ritwik Bhaduri and Aabesh Bhattacharyya and Rina Foygel Barber and Lucas Janson},
  year={2025},
  eprint={2511.05281},
  archivePrefix={arXiv},
  primaryClass={stat.ME},
  url={https://arxiv.org/abs/2511.05281},
  abstract={Tests of goodness of fit are used in nearly every domain where statistics is applied. One powerful and flexible approach is to sample artificial data sets that are exchangeable with the real data under the null hypothesis (but not under the alternative), as this allows the analyst to conduct a valid test using any test statistic they desire. Such sampling is typically done by conditioning on either an exact or approximate sufficient statistic, but existing methods for doing so have significant limitations, which either preclude their use or substantially reduce their power or computational tractability for many important models. In this paper, we propose to condition on samples from a Bayesian posterior distribution, which constitute a very different type of approximate sufficient statistic than those considered in prior work. Our approach, approximately co-sufficient sampling via Bayes (aCSS-B), considerably expands the scope of this flexible type of goodness-of-fit testing. We prove the approximate validity of the resulting test, and demonstrate its utility on three common null models where no existing methods apply, as well as its outperformance on models where existing methods do apply.}
}

@misc{bhaduri2024compositional,
  title={Compositional Covariate Importance Testing via Partial Conjunction of Bivariate Hypotheses},
  author={Ritwik Bhaduri and Siyuan Ma and Lucas Janson},
  year={2024},
  eprint={2501.00566},
  archivePrefix={arXiv},
  primaryClass={stat.ME},
  url={https://arxiv.org/abs/2501.00566},
  abstract={Compositional data (i.e., data comprising random variables that sum up to a constant) arises in many applications including microbiome studies, chemical ecology, political science, and experimental designs. Yet when compositional data serve as covariates in a regression, the sum constraint renders every covariate automatically conditionally independent of the response given the other covariates, since each covariate is a deterministic function of the others. Since essentially all covariate importance tests and variable selection methods, including parametric ones, are at their core testing conditional independence, they are all completely powerless on regression problems with compositional covariates. In fact, compositionality causes ambiguity in the very notion of relevant covariates. To address this problem, we identify a natural way to translate the typical notion of relevant covariates to the setting with compositional covariates and establish that it is intuitive, well-defined, and unique. We then develop corresponding hypothesis tests and controlled variable selection procedures via a novel connection with bivariate conditional independence testing and partial conjunction hypothesis testing. Finally, we provide theoretical guarantees of the validity of our methods, and through numerical experiments demonstrate that our methods are not only valid but also powerful across a range of data-generating scenarios.}
}

@article{drozario2023circulating,
  title={Circulating interleukin-8 dynamics parallels disease course and is linked to clinical outcomes in severe COVID-19},
  author={Ranit D'Rozario and Deblina Raychaudhuri and Purbita Bandopadhyay and Jafar Sarif and Priyanka Mehta and Chinky Shiu Chen Liu and Bishnu Prasad Sinha and Jayasree Roy and Ritwik Bhaduri and Monidipa Das and others},
  journal={Viruses},
  year={2023},
  abstract={Severe COVID-19 frequently features a systemic deluge of cytokines. Circulating cytokines that can stratify risks are useful for more effective triage and management. Here, we ran a machine-learning algorithm on a dataset of 36 plasma cytokines in a cohort of severe COVID-19 to identify cytokine/s useful for describing the dynamic clinical state in multiple regression analysis. We performed RNA-sequencing of circulating blood cells collected at different time-points. From a Bayesian Information Criterion analysis, a combination of interleukin-8 (IL-8), Eotaxin, and Interferon-gamma (IFNgamma) was found to be significantly linked to blood oxygenation over seven days. Individually testing the cytokines in receiver operator characteristics analyses identified IL-8 as a strong stratifier for clinical outcomes. Circulating IL-8 dynamics paralleled disease course. We also revealed key transitions in immune transcriptome in patients stratified for circulating IL-8 at three time-points. The study identifies plasma IL-8 as a key pathogenic cytokine linking systemic hyper-inflammation to the clinical outcomes in COVID-19.}
}

@article{ray2022phase2,
  title={A phase 2 single center open label randomised control trial for convalescent plasma therapy in patients with severe COVID-19},
  author={Yogiraj Ray and Shekhar Ranjan Paul and Purbita Bandopadhyay and Ranit D'Rozario and Jafar Sarif and Deblina Raychaudhuri and Debaleena Bhowmik and Abhishake Lahiri and Janani Srinivasa Vasudevan and Ranjeet Maurya and others},
  journal={Nature Communications},
  year={2022},
  doi={10.1038/s41467-022-28064-7},
  abstract={A single center open label phase 2 randomised control trial (Clinical Trial Registry of India No. CTRI/2020/05/025209) was done to assess clinical and immunological benefits of passive immunization using convalescent plasma therapy. At the Infectious Diseases and Beleghata General Hospital in Kolkata, India, 80 patients hospitalized with severe COVID-19 disease were recruited and randomized into either standard of care (SOC, N = 40) or convalescent plasma therapy (CPT, N = 40). Primary outcomes were all-cause mortality by day 30 of enrolment and immunological correlates of response to therapy. The trial found that all-cause mortality was not significantly different among severe COVID-19 patients with ARDS randomized to the two treatment arms (Mantel-Haenszel Hazard Ratio 0.6731, 95% CI 0.3010-1.505, P = 0.3424). No adverse effect was reported with CPT. In severe COVID-19 patients with mild or moderate ARDS no significant clinical benefit was registered in this clinical trial with convalescent plasma therapy in terms of prespecified outcomes.}
}

@article{raychaudhuri2022clinical,
  title={Clinical Trial Subgroup Analyses to Investigate Clinical and Immunological Outcomes of Convalescent Plasma Therapy in Severe COVID-19},
  author={Deblina Raychaudhuri and Purbita Bandopadhyay and Ranit D'Rozario and Jafar Sarif and Yogiraj Ray and Shekhar Ranjan Paul and Praveen Singh and Kausik Chaudhuri and Ritwik Bhaduri and Rajesh Pandey and others},
  journal={Mayo Clinic Proceedings: Innovations, Quality & Outcomes},
  year={2022},
  doi={10.1016/j.mayocpiqo.2022.09.001},
  abstract={A series of subclass analyses were performed on previously published outcome data and accompanying clinical metadata from a completed randomized controlled trial assessing clinical and immunological benefits of convalescent plasma therapy (CPT). Although the primary clinical outcomes were not significantly different in the RCT across all age groups, significant immediate mitigation of hypoxia, reduction in hospital stay, and significant survival benefit were registered in younger patients with severe COVID-19 and acute respiratory distress syndrome on receiving CPT. Subgroup analyses revealed that clinical benefits of CPT in severe COVID-19 are linked to the anti-inflammatory protein content of convalescent plasma apart from the anti-SARS-CoV-2 neutralizing antibody content.}
}

@article{bhaduri2022seirfansy,
  title={Extending the susceptible-exposed-infected-removed ({SEIR}) model to handle the false negative rate and symptom-based administration of {COVID}-19 diagnostic tests: {SEIR-fansy}},
  author={Ritwik Bhaduri and Ritoban Kundu and Soumik Purkayastha and Michael Kleinsasser and Lauren J. Beesley and Bhramar Mukherjee and others},
  journal={Statistics in Medicine},
  year={2022},
  doi={10.1002/sim.9357},
  abstract={False negative rates of diagnostic tests for SARS-CoV-2, together with selection bias due to prioritized testing, can result in inaccurate modeling of COVID-19 transmission dynamics based on reported case counts. We propose an extension of the widely used SEIR model that accounts for misclassification error and selection bias, and derive an analytic expression for the basic reproduction number R0 as a function of false negative rates of the diagnostic tests and selection probabilities for getting tested. Analyzing data from the first two waves of the pandemic in India, we show that correcting for misclassification and selection leads to more accurate prediction in a test sample. An R-package SEIRfansy is developed for broader dissemination.}
}

@article{bhaduri2022roughfuzzy,
  title={Rough-Fuzzy {CPD}: a gradual change point detection algorithm},
  author={Ritwik Bhaduri and Subhrajyoty Roy and Sankar K. Pal},
  journal={Journal of Data, Information and Management},
  year={2022},
  doi={10.1007/s42488-022-00077-3},
  abstract={Changepoint detection is the problem of finding abrupt or gradual changes in time series data when the distribution of the time series changes significantly. Here we present a new approach to solve changepoint detection using fuzzy rough set theory which is able to detect such gradual changepoints. An expression for the rough-fuzzy estimate of changepoints is derived along with its mathematical properties concerning fast computation. Simulation studies reveal that the proposed method beats other fuzzy methods and also popular crisp methods like WBS, PELT and BOCD in detecting gradual changepoints. We have developed the Python package 'roufcp' for broader dissemination of the methods.}
}

@article{purkayastha2021comparison,
  title={A comparison of five epidemiological models for transmission of {SARS-CoV-2} in India},
  author={Soumik Purkayastha and Rupam Bhattacharyya and Ritwik Bhaduri and Ritoban Kundu and Xuelin Gu and Maxwell Salvatore and Debashree Ray and Swapnil Mishra and others},
  journal={BMC Infectious Diseases},
  year={2021},
  doi={10.1186/s12879-021-06077-9},
  abstract={We study how five epidemiological models forecast and assess the course of the COVID-19 pandemic in India: a baseline curve-fitting model, an extended SIR (eSIR) model, two extended SEIR (SAPHIRE and SEIR-fansy) models, and a semi-mechanistic Bayesian hierarchical model (ICM). Using COVID-19 case-recovery-death count data reported in India from March 15 to October 15, 2020, we generate predictions and compare prediction accuracy. For cumulative case counts, SMAPE values range from 2.25% (SAPHIRE) to 6.89% (baseline). Overall, the SEIR-fansy model appeared to be a good choice with a publicly available R-package and desired flexibility plus accuracy.}
}

@article{babu2021covid,
  title={{COVID}-19 pandemic in India: Through the lens of modeling},
  author={Giridhara R. Babu and Debashree Ray and Ritwik Bhaduri and Aritra Halder and Ritoban Kundu and Gautam I. Menon and others},
  journal={Global Health: Science and Practice},
  year={2021},
  doi={10.9745/GHSP-D-21-00233},
  abstract={India has devised innovative strategies to reduce the spread of COVID-19 within the constraints of a low-resource setting, while also making some questionable policy decisions. In this commentary, as a team of public health data scientists engaged in modeling the pandemic since early 2020, we reflect on India's journey over the past year, examining the landscape of epidemiological models, the challenges of mismeasured case and death counts, and lessons for public health and data infrastructure that can be globally valuable.}
}

@misc{mukherjee2021estimating,
  title={Estimating the infection fatality rate from {SARS-CoV-2} in India},
  author={Bhramar Mukherjee and Soumik Purkayashtha and Ritoban Kundu and Ritwik Bhaduri and others},
  year={2021},
  note={Available at SSRN 3798552},
  doi={10.2139/ssrn.3798552},
  abstract={There has been much discussion and debate around the underreporting of COVID-19 infections and deaths in India. We estimate the underreporting factor for infections from national seroprevalence surveys and use a rigorous compartmental epidemiologic model to estimate the undetected number of infections and deaths. Model-based estimated underreporting factor for infections is 11.11 (95% CI 10.71-11.47) and for deaths is 3.56 (95% CI 3.48-3.64), implying approximately 91% of infections and 72% of deaths remain unreported. The infection fatality rate for India is estimated at 0.13% based on reported deaths and 0.46% when accounting for death underreporting.}
}

@article{purkayastha2021estimating,
  title={Estimating the wave 1 and wave 2 infection fatality rates from {SARS-CoV-2} in India},
  author={Soumik Purkayastha and Ritoban Kundu and Ritwik Bhaduri and Daniel Barker and Michael Kleinsasser and Debashree Ray and others},
  journal={BMC Research Notes},
  year={2021},
  doi={10.1186/s13104-021-05652-2},
  abstract={We estimate the underreporting factor for COVID-19 infections and deaths in India using a compartmental epidemiologic model across two waves: wave 1 (April 1, 2020-January 31, 2021) and part of wave 2 (February 1-May 15, 2021). Model-based estimated underreporting factor for infections is 11.11 (95% CrI 10.71-11.47) for wave 1 and 26.77 (95% CrI 24.26-28.81) for wave 2. Combining waves 1 and 2, estimated total infections stand at 491 million (36% of the population) with a combined IFR of 0.25%.}
}

@article{bhattacharyya2021incorporating,
  title={Incorporating false negative tests in epidemiological models for {SARS-CoV-2} transmission and reconciling with seroprevalence estimates},
  author={Rupam Bhattacharyya and Ritoban Kundu and Ritwik Bhaduri and Debashree Ray and Lauren J. Beesley and Maxwell Salvatore and Bhramar Mukherjee},
  journal={Scientific Reports},
  year={2021},
  doi={10.1038/s41598-021-89127-1},
  abstract={SEIR-type epidemiologic models can predict unreported cases and deaths assuming perfect testing. We apply a method to account for the high false negative rates of diagnostic RT-PCR tests for SARS-CoV-2 in a classic SEIR model and validate projections against population-based serosurveys for Delhi, India. Applying our method to training data from Delhi (March 15-June 30, 2020), we estimate the underreporting factor for cases at 34-53 (deaths: 8-13), consistent with the first round of serosurveys with 22.86% IgG antibody prevalence. Such model-based estimates provide a viable alternative to resource-intensive serosurveys for tracking unreported cases and deaths.}
}

@article{zimmermann2021sars,
  title={{SARS-CoV-2} infection fatality rates in India: systematic review, meta-analysis and model-based estimation},
  author={Lauren Zimmermann and Subarna Bhattacharya and Soumik Purkayastha and Ritoban Kundu and Ritwik Bhaduri and Parikshit Ghosh and others},
  journal={Studies in Microeconomics},
  year={2021},
  doi={10.1177/23210222211054324},
  abstract={We synthesize the existing literature on SARS-CoV-2 infection fatality rates (IFR) in India through systematic review and meta-analysis. Nationwide pooled IFR estimates are 0.097% (95% CI: 0.067-0.140) based on reported deaths and 0.365%-0.485% when accounting for death underreporting. Underreporting factors for cases range from 14.3-29.1 across four nationwide serosurveys; for deaths they range from 4.4-11.9, with cumulative excess deaths of 1.79-4.9 million as of June 2021. Updated SEIR model-based estimates largely reconcile with the empirical findings.}
}

@misc{bhattacharyya2020reconciling,
  title={Reconciling epidemiological models with misclassified case-counts for {SARS-CoV-2} with seroprevalence surveys: a case study in Delhi, India},
  author={Rupam Bhattacharyya and Ritwik Bhaduri and Ritoban Kundu and Maxwell Salvatore and others},
  year={2020},
  howpublished={medRxiv},
  doi={10.1101/2020.07.31.20166249},
  abstract={Underreporting of COVID-19 cases and deaths is a hindrance to correctly modeling and monitoring the pandemic. We develop a method to account for high false negative rates in RT-PCR in an extension to the classic SEIR model, applying it to Delhi, India. We obtain estimates of the underreporting factor for cases at 34-53 times and for deaths at 8-13 times, largely consistent with the first round of serosurveys for Delhi with an estimated 22.86% seroprevalence, implying approximately 96-98% of cases in Delhi remained unreported.}
}

@misc{bhaduri2019onset,
  title={Onset detection: A new approach to {QBH} system},
  author={Ritwik Bhaduri and Soham Bonnerjee and Subhrajyoty Roy},
  year={2019},
  eprint={1908.07409},
  archivePrefix={arXiv},
  primaryClass={stat.AP},
  url={https://arxiv.org/abs/1908.07409},
  abstract={Query by Humming (QBH) is a system to provide a user with the song(s) which the user hums to the system. Current QBH methods require the extraction of onset and pitch information in order to track similarity with various versions of different songs. We focus on detecting precise onsets only and use them to build a QBH system which is better than existing methods in terms of speed and memory, and empirically in terms of accuracy. We also provide statistical analogy for onset detection functions and provide a measure of error in our algorithm.}
}