Clinical risk prediction models are commonly developed in a post-hoc and passive fashion, capitalizing on convenient data from completed clinical trials or retrospective cohorts. Impacts of the models often end at their publication rather than with the patients. The field of clinical risk prediction is rapidly improving in a progressively more transparent data science era. Based on collective experience over the past decade by the Prostate Biopsy Collaborative Group (PBCG), this paper proposes the following four data science-driven strategies for improving clinical risk prediction to the benefit of clinical practice and research. The first proposed strategy is to actively design prospective data collection, monitoring, analysis and validation of risk tools following the same standards as for clinical trials in order to elevate the quality of training data. The second suggestion is to make risk tools and model formulas available online. User-friendly risk tools will bring quantitative information to patients and their clinicians for improved knowledge-based decision-making. As past experience testifies, online tools expedite independent validation, providing helpful information as to whether the tools are generalizable to new populations. The third proposal is to dynamically update and localize risk tools to adapt to changing demographic and clinical landscapes. The fourth strategy is to accommodate systematic missing data patterns across cohorts in order to maximize the statistical power in model training, as well as to accommodate missing information on the end-user side too, in order to maximize utility for the public.
{"title":"Active Data Science for Improving Clinical Risk Prediction","authors":"D. Ankerst, Matthias Neumair","doi":"10.6339/22-jds1078","DOIUrl":"https://doi.org/10.6339/22-jds1078","url":null,"abstract":"Clinical risk prediction models are commonly developed in a post-hoc and passive fashion, capitalizing on convenient data from completed clinical trials or retrospective cohorts. Impacts of the models often end at their publication rather than with the patients. The field of clinical risk prediction is rapidly improving in a progressively more transparent data science era. Based on collective experience over the past decade by the Prostate Biopsy Collaborative Group (PBCG), this paper proposes the following four data science-driven strategies for improving clinical risk prediction to the benefit of clinical practice and research. The first proposed strategy is to actively design prospective data collection, monitoring, analysis and validation of risk tools following the same standards as for clinical trials in order to elevate the quality of training data. The second suggestion is to make risk tools and model formulas available online. User-friendly risk tools will bring quantitative information to patients and their clinicians for improved knowledge-based decision-making. As past experience testifies, online tools expedite independent validation, providing helpful information as to whether the tools are generalizable to new populations. The third proposal is to dynamically update and localize risk tools to adapt to changing demographic and clinical landscapes. The fourth strategy is to accommodate systematic missing data patterns across cohorts in order to maximize the statistical power in model training, as well as to accommodate missing information on the end-user side too, in order to maximize utility for the public.","PeriodicalId":73699,"journal":{"name":"Journal of data science : JDS","volume":"1 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2022-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"71320795","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
How do statistical regression results compare to intuitive, visually fitted results? Fitting lines by eye through a set of points has been explored since the 20th century. Common methods of fitting trends by eye involve maneuvering a string, black thread, or ruler until the fit is suitable, then drawing the line through the set of points. In 2015, the New York Times introduced an interactive feature, called ‘You Draw It,’ where readers are asked to input their own assumptions about various metrics and compare how these assumptions relate to reality. This research is intended to implement ‘You Draw It’, adapted from the New York Times, as a way to measure the patterns we see in data. In this paper, we describe the adaptation of an old tool for graphical testing and evaluation, eye-fitting, for use in modern web-applications suitable for testing statistical graphics. We present an empirical evaluation of this testing method for linear regression, and briefly discuss an extension of this method to non-linear applications.
{"title":"‘You Draw It’: Implementation of Visually Fitted Trends with r2d3","authors":"Emily A. Robinson, Réka Howard, Susan Vanderplas","doi":"10.6339/22-jds1083","DOIUrl":"https://doi.org/10.6339/22-jds1083","url":null,"abstract":"How do statistical regression results compare to intuitive, visually fitted results? Fitting lines by eye through a set of points has been explored since the 20th century. Common methods of fitting trends by eye involve maneuvering a string, black thread, or ruler until the fit is suitable, then drawing the line through the set of points. In 2015, the New York Times introduced an interactive feature, called ‘You Draw It,’ where readers are asked to input their own assumptions about various metrics and compare how these assumptions relate to reality. This research is intended to implement ‘You Draw It’, adapted from the New York Times, as a way to measure the patterns we see in data. In this paper, we describe the adaptation of an old tool for graphical testing and evaluation, eye-fitting, for use in modern web-applications suitable for testing statistical graphics. We present an empirical evaluation of this testing method for linear regression, and briefly discuss an extension of this method to non-linear applications.","PeriodicalId":73699,"journal":{"name":"Journal of data science : JDS","volume":"1 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2022-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"71320862","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Jiabu Ye, Binbing Yu, H. Mann, A. Sabin, Z. Szíjgyártó, David Wright, P. Mukhopadhyay, C. Massacesi, S. Ghiorghiu, R. Iacona
We assessed the impact of the coronavirus disease 2019 (COVID-19) pandemic on the statistical analysis of time-to-event outcomes in late-phase oncology trials. Using a simulated case study that mimics a Phase III ongoing trial during the pandemic, we evaluated the impact of COVID-19-related deaths, time off-treatment and missed clinical visits due to the pandemic, on overall survival and/or progression-free survival in terms of test size (also referred to as Type 1 error rate or alpha level), power, and hazard ratio (HR) estimates. We found that COVID-19-related deaths would impact both size and power, and lead to biased HR estimates; the impact would be more severe if there was an imbalance in COVID-19-related deaths between the study arms. Approaches censoring COVID-19-related deaths may mitigate the impact on power and HR estimation, especially if study data cut-off was extended to recover censoring-related event loss. The impact of COVID-19-related time off-treatment would be modest for power, and moderate for size and HR estimation. Different rules of censoring cancer progression times result in a slight difference in the power for the analysis of progression-free survival. The simulations provided valuable information for determining whether clinical-trial modifications should be required for ongoing trials during the COVID-19 pandemic.
{"title":"Addressing the Impact of the COVID-19 Pandemic on Survival Outcomes in Randomized Phase III Oncology Trials","authors":"Jiabu Ye, Binbing Yu, H. Mann, A. Sabin, Z. Szíjgyártó, David Wright, P. Mukhopadhyay, C. Massacesi, S. Ghiorghiu, R. Iacona","doi":"10.6339/22-jds1079","DOIUrl":"https://doi.org/10.6339/22-jds1079","url":null,"abstract":"We assessed the impact of the coronavirus disease 2019 (COVID-19) pandemic on the statistical analysis of time-to-event outcomes in late-phase oncology trials. Using a simulated case study that mimics a Phase III ongoing trial during the pandemic, we evaluated the impact of COVID-19-related deaths, time off-treatment and missed clinical visits due to the pandemic, on overall survival and/or progression-free survival in terms of test size (also referred to as Type 1 error rate or alpha level), power, and hazard ratio (HR) estimates. We found that COVID-19-related deaths would impact both size and power, and lead to biased HR estimates; the impact would be more severe if there was an imbalance in COVID-19-related deaths between the study arms. Approaches censoring COVID-19-related deaths may mitigate the impact on power and HR estimation, especially if study data cut-off was extended to recover censoring-related event loss. The impact of COVID-19-related time off-treatment would be modest for power, and moderate for size and HR estimation. Different rules of censoring cancer progression times result in a slight difference in the power for the analysis of progression-free survival. The simulations provided valuable information for determining whether clinical-trial modifications should be required for ongoing trials during the COVID-19 pandemic.","PeriodicalId":73699,"journal":{"name":"Journal of data science : JDS","volume":"1 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2022-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"71320957","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Many undergraduate students who matriculated in Science, Technology, Engineering and Mathematics (STEM) degree programs drop out or switch their major. Previous studies indicate that performance of students in prerequisite courses is important for attrition of students in STEM. This study analyzed demographic information, ACT/SAT score, and performance of students in freshman year courses to develop machine learning models predicting their success in earning a bachelor’s degree in biology. The predictive model based on Random Forest (RF) and Extreme Gradient Boosting (XGBoost) showed a better performance in terms of AUC (Area Under the Curve) with more balanced sensitivity and specificity than Logistic Regression (LR), K-Nearest Neighbor (KNN), and Neural Network (NN) models. An explainable machine learning approach called break-down was employed to identify important freshman year courses that could have a larger impact on student success at the biology degree program and student levels. More important courses identified at the program level can help program coordinators to prioritize their effort in addressing student attrition while more important courses identified at the student level can help academic advisors to provide more personalized, data-driven guidance to students.
{"title":"Identifying Prerequisite Courses in Undergraduate Biology Using Machine Learning","authors":"Youngjin Lee","doi":"10.6339/22-jds1068","DOIUrl":"https://doi.org/10.6339/22-jds1068","url":null,"abstract":"Many undergraduate students who matriculated in Science, Technology, Engineering and Mathematics (STEM) degree programs drop out or switch their major. Previous studies indicate that performance of students in prerequisite courses is important for attrition of students in STEM. This study analyzed demographic information, ACT/SAT score, and performance of students in freshman year courses to develop machine learning models predicting their success in earning a bachelor’s degree in biology. The predictive model based on Random Forest (RF) and Extreme Gradient Boosting (XGBoost) showed a better performance in terms of AUC (Area Under the Curve) with more balanced sensitivity and specificity than Logistic Regression (LR), K-Nearest Neighbor (KNN), and Neural Network (NN) models. An explainable machine learning approach called break-down was employed to identify important freshman year courses that could have a larger impact on student success at the biology degree program and student levels. More important courses identified at the program level can help program coordinators to prioritize their effort in addressing student attrition while more important courses identified at the student level can help academic advisors to provide more personalized, data-driven guidance to students.","PeriodicalId":73699,"journal":{"name":"Journal of data science : JDS","volume":"491 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2022-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"71320701","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Daeyoung Lim, Ming-Hui Chen, N. Ravishanker, Mark Bolduc, Brian McKeon, Stanley Nolan
The complexity of energy infrastructure at large institutions increasingly calls for data-driven monitoring of energy usage. This article presents a hybrid monitoring algorithm for detecting consumption surges using statistical hypothesis testing, leveraging the posterior distribution and its information about uncertainty to introduce randomness in the parameter estimates, while retaining the frequentist testing framework. This hybrid approach is designed to be asymptotically equivalent to the Neyman-Pearson test. We show via extensive simulation studies that the hybrid approach enjoys control over type-1 error rate even with finite sample sizes whereas the naive plug-in method tends to exceed the specified level, resulting in overpowered tests. The proposed method is applied to the natural gas usage data at the University of Connecticut.
{"title":"A Hybrid Monitoring Procedure for Detecting Abnormality with Application to Energy Consumption Data","authors":"Daeyoung Lim, Ming-Hui Chen, N. Ravishanker, Mark Bolduc, Brian McKeon, Stanley Nolan","doi":"10.6339/22-jds1039","DOIUrl":"https://doi.org/10.6339/22-jds1039","url":null,"abstract":"The complexity of energy infrastructure at large institutions increasingly calls for data-driven monitoring of energy usage. This article presents a hybrid monitoring algorithm for detecting consumption surges using statistical hypothesis testing, leveraging the posterior distribution and its information about uncertainty to introduce randomness in the parameter estimates, while retaining the frequentist testing framework. This hybrid approach is designed to be asymptotically equivalent to the Neyman-Pearson test. We show via extensive simulation studies that the hybrid approach enjoys control over type-1 error rate even with finite sample sizes whereas the naive plug-in method tends to exceed the specified level, resulting in overpowered tests. The proposed method is applied to the natural gas usage data at the University of Connecticut.","PeriodicalId":73699,"journal":{"name":"Journal of data science : JDS","volume":"1 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2022-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"71320187","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Law and legal studies has been an exciting new field for data science applications whereas the technological advancement also has profound implications for legal practice. For example, the legal industry has accumulated a rich body of high quality texts, images and other digitised formats, which are ready to be further processed and analysed by data scientists. On the other hand, the increasing popularity of data science has been a genuine challenge to legal practitioners, regulators and even general public and has motivated a long-lasting debate in the academia focusing on issues such as privacy protection and algorithmic discrimination. This paper collects 1236 journal articles involving both law and data science from the platform Web of Science to understand the patterns and trends of this interdisciplinary research field in terms of English journal publications. We find a clear trend of increasing publication volume over time and a strong presence of high-impact law and political science journals. We then use the Latent Dirichlet Allocation (LDA) as a topic modelling method to classify the abstracts into four topics based on the coherence measure. The four topics identified confirm that both challenges and opportunities have been investigated in this interdisciplinary field and help offer directions for future research.
法律和法律研究一直是数据科学应用的一个令人兴奋的新领域，而技术的进步也对法律实践产生了深远的影响。例如，法律行业已经积累了丰富的高质量文本、图像和其他数字化格式，可供数据科学家进一步处理和分析。另一方面，数据科学的日益普及对法律从业者、监管机构甚至公众都是一个真正的挑战，并引发了学术界长期以来围绕隐私保护和算法歧视等问题的辩论。本文从Web of science平台上收集了1236篇涉及法律和数据科学的期刊文章，从英文期刊发表的角度来了解这一跨学科研究领域的模式和趋势。我们发现，随着时间的推移，出版物数量有明显的增长趋势，高影响力的法律和政治学期刊也有很强的影响力。然后，我们使用潜在狄利克雷分配(Latent Dirichlet Allocation, LDA)作为主题建模方法，根据一致性度量将摘要分为四个主题。确定的四个主题确认了这一跨学科领域的挑战和机遇，并有助于为未来的研究提供方向。
{"title":"Data Science Applications and Implications in Legal Studies: A Perspective Through Topic Modelling","authors":"Jinzhe Tan, Huan Wan, Ping Yan, Zheng Hua Zhu","doi":"10.6339/22-jds1058","DOIUrl":"https://doi.org/10.6339/22-jds1058","url":null,"abstract":"Law and legal studies has been an exciting new field for data science applications whereas the technological advancement also has profound implications for legal practice. For example, the legal industry has accumulated a rich body of high quality texts, images and other digitised formats, which are ready to be further processed and analysed by data scientists. On the other hand, the increasing popularity of data science has been a genuine challenge to legal practitioners, regulators and even general public and has motivated a long-lasting debate in the academia focusing on issues such as privacy protection and algorithmic discrimination. This paper collects 1236 journal articles involving both law and data science from the platform Web of Science to understand the patterns and trends of this interdisciplinary research field in terms of English journal publications. We find a clear trend of increasing publication volume over time and a strong presence of high-impact law and political science journals. We then use the Latent Dirichlet Allocation (LDA) as a topic modelling method to classify the abstracts into four topics based on the coherence measure. The four topics identified confirm that both challenges and opportunities have been investigated in this interdisciplinary field and help offer directions for future research.","PeriodicalId":73699,"journal":{"name":"Journal of data science : JDS","volume":"1 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2022-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"71320450","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This special issue features eight articles on “Large-Scale Spatial Data Science.” Data science for complex and large-scale spatial and spatio-temporal data has become essential in many research fields, such as climate science and environmental applications. Due to the ever-increasing amounts of data collected, traditional statistical approaches tend to break down and computa-tionally efficient methods and scalable algorithms that are suitable for large-scale spatial data have become crucial to cope with many challenges associated with big data. This special issue aims at highlighting some of the latest developments in the area of large-scale spatial data science. The research papers presented showcase advanced statistical methods and machine learn-ing approaches for solving complex and large-scale problems arising from modern data science applications. Abdulah et al. (2022) reported the results of the second competition on spatial statistics for large datasets organized by the King Abdullah University of Science and Technology (KAUST). Very large datasets (up to 1 million in size) were generated with the ExaGeoStat software to design the competition on large-scale predictions in challenging settings, including univariate nonstationary spatial processes, univariate stationary space-time processes, and bivariate stationary spatial processes. The authors described the data generation process in detail in each setting and made these valuable datasets publicly available. They reviewed the methods used by fourteen competing teams worldwide, analyzed the results of the competition, and assessed the performance of each team.
{"title":"Editorial: Large-Scale Spatial Data Science","authors":"Sameh Abdulah, S. Castruccio, M. Genton, Ying Sun","doi":"10.6339/22-jds204edi","DOIUrl":"https://doi.org/10.6339/22-jds204edi","url":null,"abstract":"This special issue features eight articles on “Large-Scale Spatial Data Science.” Data science for complex and large-scale spatial and spatio-temporal data has become essential in many research fields, such as climate science and environmental applications. Due to the ever-increasing amounts of data collected, traditional statistical approaches tend to break down and computa-tionally efficient methods and scalable algorithms that are suitable for large-scale spatial data have become crucial to cope with many challenges associated with big data. This special issue aims at highlighting some of the latest developments in the area of large-scale spatial data science. The research papers presented showcase advanced statistical methods and machine learn-ing approaches for solving complex and large-scale problems arising from modern data science applications. Abdulah et al. (2022) reported the results of the second competition on spatial statistics for large datasets organized by the King Abdullah University of Science and Technology (KAUST). Very large datasets (up to 1 million in size) were generated with the ExaGeoStat software to design the competition on large-scale predictions in challenging settings, including univariate nonstationary spatial processes, univariate stationary space-time processes, and bivariate stationary spatial processes. The authors described the data generation process in detail in each setting and made these valuable datasets publicly available. They reviewed the methods used by fourteen competing teams worldwide, analyzed the results of the competition, and assessed the performance of each team.","PeriodicalId":73699,"journal":{"name":"Journal of data science : JDS","volume":"9 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2022-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"71320563","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The 2020 Census County Assessment Tool was developed to assist decennial census data users in identifying deviations between expected census counts and the released counts across population and housing indicators. The tool also offers contextual data for each county on factors which could have contributed to census collection issues, such as self-response rates and COVID-19 infection rates. The tool compiles this information into a downloadable report and includes additional local data sources relevant to the data collection process and experts to seek more assistance.
{"title":"Creating a Census County Assessment Tool for Visualizing Census Data","authors":"Izzy Youngs, R. Prevost, Christopher Dick","doi":"10.6339/22-jds1082","DOIUrl":"https://doi.org/10.6339/22-jds1082","url":null,"abstract":"The 2020 Census County Assessment Tool was developed to assist decennial census data users in identifying deviations between expected census counts and the released counts across population and housing indicators. The tool also offers contextual data for each county on factors which could have contributed to census collection issues, such as self-response rates and COVID-19 infection rates. The tool compiles this information into a downloadable report and includes additional local data sources relevant to the data collection process and experts to seek more assistance.","PeriodicalId":73699,"journal":{"name":"Journal of data science : JDS","volume":"1 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2022-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"71320850","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Many small and rural places are shrinking. Interactive dashboards are the most common use cases for data visualization and context for exploratory data tools. In our paper, we will use Iowa data to explore the specific scope of how dashboards are used in small and rural area to empower novice analysts to make data-driven decisions. Our framework will suggest a number of research directions to better support small and rural places from shrinking using an interactive dashboard design, implementation and use for the every day analyst.
{"title":"Exploring Rural Shrink Smart Through Guided Discovery Dashboards","authors":"Denise Bradford, Susan Vanderplas","doi":"10.6339/22-jds1080","DOIUrl":"https://doi.org/10.6339/22-jds1080","url":null,"abstract":"Many small and rural places are shrinking. Interactive dashboards are the most common use cases for data visualization and context for exploratory data tools. In our paper, we will use Iowa data to explore the specific scope of how dashboards are used in small and rural area to empower novice analysts to make data-driven decisions. Our framework will suggest a number of research directions to better support small and rural places from shrinking using an interactive dashboard design, implementation and use for the every day analyst.","PeriodicalId":73699,"journal":{"name":"Journal of data science : JDS","volume":"1 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2022-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"71320966","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
When releasing data to the public, a vital concern is the risk of exposing personal information of the individuals who have contributed to the data set. Many mechanisms have been proposed to protect individual privacy, though less attention has been dedicated to practically conducting valid inferences on the altered privacy-protected data sets. For frequency tables, the privacy-protection-oriented perturbations often lead to negative cell counts. Releasing such tables can undermine users’ confidence in the usefulness of such data sets. This paper focuses on releasing one-way frequency tables. We recommend an optimal mechanism that satisfies ϵ-differential privacy (DP) without suffering from having negative cell counts. The procedure is optimal in the sense that the expected utility is maximized under a given privacy constraint. Valid inference procedures for testing goodness-of-fit are also developed for the DP privacy-protected data. In particular, we propose a de-biased test statistic for the optimal procedure and derive its asymptotic distribution. In addition, we also introduce testing procedures for the commonly used Laplace and Gaussian mechanisms, which provide a good finite sample approximation for the null distributions. Moreover, the decaying rate requirements for the privacy regime are provided for the inference procedures to be valid. We further consider common users’ practices such as merging related or neighboring cells or integrating statistical information obtained across different data sources and derive valid testing procedures when these operations occur. Simulation studies show that our inference results hold well even when the sample size is relatively small. Comparisons with the current field standards, including the Laplace, the Gaussian (both with/without post-processing of replacing negative cell counts with zeros), and the Binomial-Beta McClure-Reiter mechanisms, are carried out. In the end, we apply our method to the National Center for Early Development and Learning’s (NCEDL) multi-state studies data to demonstrate its practical applicability.
{"title":"Inference for Optimal Differential Privacy Procedures for Frequency Tables","authors":"Chengcheng Li, Na Wang, Gongjun Xu","doi":"10.6339/22-jds1044","DOIUrl":"https://doi.org/10.6339/22-jds1044","url":null,"abstract":"When releasing data to the public, a vital concern is the risk of exposing personal information of the individuals who have contributed to the data set. Many mechanisms have been proposed to protect individual privacy, though less attention has been dedicated to practically conducting valid inferences on the altered privacy-protected data sets. For frequency tables, the privacy-protection-oriented perturbations often lead to negative cell counts. Releasing such tables can undermine users’ confidence in the usefulness of such data sets. This paper focuses on releasing one-way frequency tables. We recommend an optimal mechanism that satisfies ϵ-differential privacy (DP) without suffering from having negative cell counts. The procedure is optimal in the sense that the expected utility is maximized under a given privacy constraint. Valid inference procedures for testing goodness-of-fit are also developed for the DP privacy-protected data. In particular, we propose a de-biased test statistic for the optimal procedure and derive its asymptotic distribution. In addition, we also introduce testing procedures for the commonly used Laplace and Gaussian mechanisms, which provide a good finite sample approximation for the null distributions. Moreover, the decaying rate requirements for the privacy regime are provided for the inference procedures to be valid. We further consider common users’ practices such as merging related or neighboring cells or integrating statistical information obtained across different data sources and derive valid testing procedures when these operations occur. Simulation studies show that our inference results hold well even when the sample size is relatively small. Comparisons with the current field standards, including the Laplace, the Gaussian (both with/without post-processing of replacing negative cell counts with zeros), and the Binomial-Beta McClure-Reiter mechanisms, are carried out. In the end, we apply our method to the National Center for Early Development and Learning’s (NCEDL) multi-state studies data to demonstrate its practical applicability.","PeriodicalId":73699,"journal":{"name":"Journal of data science : JDS","volume":"1 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2022-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"71320068","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：