When Costs Are Unequal and Unknown: A Subtree Grafting Approach for Unbalanced Data Classification*

In binary classifications, a decision tree learned from unbalanced data typically creates an important challenge related to the high misclassification rate of the minority class. Assigning different misclassification costs can address this problem, though usually at the cost of accuracy for the majority class. This effect can be particularly hazardous if the costs cannot be specified precisely. When the costs are unknown or difficult to determine, decision makers may prefer a classifier with more balanced accuracy for both classes rather than a standard or cost‐sensitively learned one. In the context of learning trees, this research therefore proposes a new tree induction approach called subtree grafting (STG). On the basis of a real bank data set and several other data sets, we test the proposed STG method and find that our proposed approach provides a successful compromise between standard and cost‐sensitive trees.     

HDDA: DataSifter: statistical obfuscation of electronic health records and other sensitive datasets

There are no practical and effective mechanisms to share high-dimensional data including sensitive information in various fields like health financial intelligence or socioeconomics without compromising either the utility of the data or exposing private personal or secure organizational information. Excessive scrambling or encoding of the information makes it less useful for modelling or analytical processing. Insufficient preprocessing may compromise sensitive information and introduce a substantial risk for re-identification of individuals by various stratification techniques. To address this problem, we developed a novel statistical obfuscation method (DataSifter) for on-the-fly de-identification of structured and unstructured sensitive high-dimensional data such as clinical data from electronic health records (EHR). DataSifter provides complete administrative control over the balance between risk of data re-identification and preservation of the data information. Simulation results suggest that DataSifter can provide privacy protection while maintaining data utility for different types of outcomes of interest. The application of DataSifter on a large autism dataset provides a realistic demonstration of its promise practical applications.     

THE CHOICE BETWEEN SETS OF REGRESSORS

This paper examines the choice of critical values for testing both non-sequential and nested sequential sets of constraints in the standard linear regression model. Modest increases in (e.g.) t-ratio critical values relative to their one-off values are often sufficient to maintain proper size. A Bayesian decision-theoretic approach, highlighted by the Schwarz (1978) criterion, provides a framework for deriving consistency and asymptotic local power properties of both forms of testing (data mining) algorithms.     

A response to M. Stone: 'How not to measure the efficiency of public services (and how one might)'

Summary.  This is a response to Stone's criticisms of the Spottiswoode report to the UK Treasury which was responding to the Treasury's request for improved methods to evaluate the efficiency and productivity of the 43 police districts in England and Wales. The Spottiswoode report recommended uses of data envelopment analysis (DEA) and stochastic frontier analysis (SFA), which Stone critiqued en route to proposing an alternative approach. Here we note some of the most serious errors in his criticism and inaccurate portrayals of DEA and SFA. Most of our attention is devoted to DEA, and to Stone's recommended alternative approach without much attention to SFA, partly because of his abbreviated discussion of the latter. In our response we attempt to be constructive as well as critical by showing how Stone's proposed approach can be joined to DEA to expand his proposal beyond limitations in his formulations.     

Strategic corporate sustainability performance of Chinese state-owned listed firms: A meta-frontier generalized directional distance function approach

Although sustainability recently became a key factor in social science, little progress has been made toward improving the measurement of sustainability performance. This paper proposes composite corporate sustainability performance indicators using a meta-frontier generalized directional distance function. This advanced approach can measure the efficiency of corporate social responsibility activities by benchmarking, while simultaneously considering industrial heterogeneities, using the meta-frontier approach. First, we propose the concept of a meta-frontier generalized directional distance function. Second, several standardized composite indicators related to corporate sustainability performance are developed. The meta-frontier directional distance function is estimated by solving a series of data envelopment analysis models. Chinese state-owned listed enterprises are then empirically examined using the proposed model. We find significant group heterogeneities in terms of corporate sustainability performance. We also derive some policy implications using the empirical results.     

Substance use disorders and referral to treatment in substantiated cases of child maltreatment

Parents with substance use disorders (SUD) require treatment and support in order to provide children with appropriate care and protection. Using the 2012 National Child Abuse and Neglect Data System (NCANDS), this paper analysed 464 313 substantiated child maltreatment reports to determine (i) the proportion and characteristics of reports involving substance abuse; and (ii) the child and caregiver/perpetrator (C/P) characteristics that predicted referral to treatment as recorded in service plans. Findings indicate that 12% (N = 53 234) of maltreatment reports involved C/P SUD. Yet, of those reports, only (19%) (N = 10 088) were referred to substance abuse treatment as part of their service plan, indicating a large gap between those who need treatment and those who receive it. This finding is important given that parental SUD is consistently linked to poorer child outcomes. Amongst other variables, reports indicate that C/P with co‐occurring emotional disturbance were three times more likely to be referred to treatment for SUD as part of service plans. Additional research is needed regarding the characteristics that distinguish C/P who receive referrals for SUD treatment in substantiated cases of child maltreatment.     

Anticipated Ethics and Regulatory Challenges in PCORnet: The National Patient-Centered Clinical Research Network

PCORnet, the National Patient-Centered Clinical Research Network, seeks to establish a robust national health data network for patient-centered comparative effectiveness research. This article reports the results of a PCORnet survey designed to identify the ethics and regulatory challenges anticipated in network implementation. A 12-item online survey was developed by leadership of the PCORnet Ethics and Regulatory Task Force; responses were collected from the 29 PCORnet networks. The most pressing ethics issues identified related to informed consent, patient engagement, privacy and confidentiality, and data sharing. High priority regulatory issues included IRB coordination, privacy and confidentiality, informed consent, and data sharing. Over 150 IRBs and five different approaches to managing multisite IRB review were identified within PCORnet. Further empirical and scholarly work, as well as practical and policy guidance, is essential if important initiatives that rely on comparative effectiveness research are to move forward.     

Perceptions concerning research integrity and the practice of data audit in the biotechnology industry

A survey was conducted in order to evaluate the responses of biotechnology executives concerning their perceptions of the frequency of questionable R & D studies performed in their companies and to assess the extent to which their companies voluntarily conducted data audits. Data audit was found to be commonly practiced on a voluntary basis by biotechnology companies, in contrast to its almost lack of practice by universities. However, public companies were more likely to practice data auditing than privately held companies. Moreover, the likelihood that managers suspected or detected questionable studies in their companies was significantly increased if the company practiced data auditing.     

Not just a number? NEETs,data and datalogical systems

This paper draws on empirical research with NEET populations (16–24-year-olds not in education, employment or training) in the U.K. in order to engage with issues around identification, data and metrics produced through datalogical systems. Our aim is to bridge contemporary discourses around data, digital bureaucracy and datalogical systems with empirical material drawn from a long-term ethnographic project with NEET groups in Leeds, U.K. in order to highlight the way datalogical systems ideologically and politically shape people’s lives. We argue that NEET is a long-standing data category that does work and has resonance within wider datalogical systems. Secondly, that these systems are decision-making and far from benign. They have real impact on people’s lives – not just in a straightforwardly, but in obscure, complex and uneven ways which makes the potential for disruption or intervention increasingly problematic. Finally, these datalogical systems also implicate and are generated by us, even as we seek to critique them.