Search

Given multiple source datasets with labels, how can we train a target model with no labeled data? Multi-source domain adaptation (MSDA) aims to train a model using multiple source datasets different from a target dataset in the absence of target data labels. MSDA is a crucial problem applicable to many practical cases where labels for the target data are unavailable due to privacy issues. Existing MSDA frameworks are limited since they align data without considering labels of the features of each domain. They also do not fully utilize the target data without labels and rely on limited feature extraction with a single extractor. In this paper, we propose Multi-EPL, a novel method for MSDA. Multi-EPL exploits label-wise moment matching to align the conditional distributions of the features for the labels, uses pseudolabels for the unavailable target labels, and introduces an ensemble of multiple feature extractors for accurate domain adaptation. Extensive experiments show that Multi-EPL provides the state-of-the-art performance for MSDA tasks in both image domains and text domains, improving the accuracy by up to 13.20%., Competing Interests: The authors have declared that no competing interests exist.

A key requirement for longitudinal studies using routinely-collected health data is to be able to measure what individuals are present in the datasets used, and over what time period. Individuals can enter and leave the covered population of administrative datasets for a variety of reasons, including both life events and characteristics of the datasets themselves. An automated, customizable method of determining individuals' presence was developed for the primary care dataset in Swansea University's SAIL Databank. The primary care dataset covers only a portion of Wales, with 76% of practices participating. The start and end date of the data varies by practice. Additionally, individuals can change practices or leave Wales. To address these issues, a two step process was developed. First, the period for which each practice had data available was calculated by measuring changes in the rate of events recorded over time. Second, the registration records for each individual were simplified. Anomalies such as short gaps and overlaps were resolved by applying a set of rules. The result of these two analyses was a cleaned set of records indicating start and end dates of available primary care data for each individual. Analysis of GP records showed that 91.0% of events occurred within periods calculated as having available data by the algorithm. 98.4% of those events were observed at the same practice of registration as that computed by the algorithm. A standardized method for solving this common problem has enabled faster development of studies using this data set. Using a rigorous, tested, standardized method of verifying presence in the study population will also positively influence the quality of research., Competing Interests: The authors have declared that no competing interests exist.

The service quality and system dependability of real-time communication networks strongly depends on the analysis of monitored data, to identify concrete problems and their causes. Many of these can be described by either their structural or temporal properties, or a combination of both. As current research is short of approaches sufficiently addressing both properties simultaneously, we propose a new feature space specifically suited for this task, which we analyze for its theoretical properties and its practical relevance. We evaluate its classification performance when used on real-world data sets of structural-temporal mobile communication data, and compare it to the performance achieved of feature representations used in related work. For this purpose we propose a system which allows the automatic detection and prediction of classes of pre-defined sequence behavior, greatly reducing costs caused by the otherwise required manual analysis. With our proposed feature spaces this system achieves a precision of more than 93% at recall values of 100%, with an up to 6.7% higher effective recall than otherwise similarly performing alternatives, notably outperforming alternative deep learning, kernel learning and ensemble learning approaches of related work. Furthermore the supported system calibration allows separating reliable from unreliable predictions more effectively, which is highly relevant for any practical application., Competing Interests: We confirm a commercial affiliation to P3 communications GmbH. This does not alter our adherence to PLOS ONE policies on sharing data and materials.

Mechanical Turk (MTurk) is a common source of research participants within the academic community. Despite MTurk's utility and benefits over traditional subject pools some researchers have questioned whether it is sustainable. Specifically, some have asked whether MTurk workers are too familiar with manipulations and measures common in the social sciences, the result of many researchers relying on the same small participant pool. Here, we show that concerns about non-naivete on MTurk are due less to the MTurk platform itself and more to the way researchers use the platform. Specifically, we find that there are at least 250,000 MTurk workers worldwide and that a large majority of US workers are new to the platform each year and therefore relatively inexperienced as research participants. We describe how inexperienced workers are excluded from studies, in part, because of the worker reputation qualifications researchers commonly use. Then, we propose and evaluate an alternative approach to sampling on MTurk that allows researchers to access inexperienced participants without sacrificing data quality. We recommend that in some cases researchers should limit the number of highly experienced workers allowed in their study by excluding these workers or by stratifying sample recruitment based on worker experience levels. We discuss the trade-offs of different sampling practices on MTurk and describe how the above sampling strategies can help researchers harness the vast and largely untapped potential of the Mechanical Turk participant pool., Competing Interests: We have read the journal’s policy and the authors of this manuscript have the following potential competing interests: All of the authors are employed at Prime Research Solutions. This is the company that owns TurkPrime, the platform whose TurkPrime ToolKit was used to source Mechanical Turk participants, and the database from which some data were queried. This does not alter our adherence to PLOS ONE policies on sharing data and materials.

Patient registry data are commonly collected as annual snapshots that need to be amalgamated to understand the longitudinal progress of each patient. However, patient identifiers can either change or may not be available for legal reasons when longitudinal data are collated from patients living in different countries. Here, we apply astronomical statistical matching techniques to link individual patient records that can be used where identifiers are absent or to validate uncertain identifiers. We adopt a Bayesian model framework used for probabilistically linking records in astronomy. We adapt this and validate it across blinded, annually collected data. This is a high-quality (Danish) sub-set of data held in the European Cystic Fibrosis Society Patient Registry (ECFSPR). Our initial experiments achieved a precision of 0.990 at a recall value of 0.987. However, detailed investigation of the discrepancies uncovered typing errors in 27 of the identifiers in the original Danish sub-set. After fixing these errors to create a new gold standard our algorithm correctly linked individual records across years achieving a precision of 0.997 at a recall value of 0.987 without recourse to identifiers. Our Bayesian framework provides the probability of whether a pair of records belong to the same patient. Unlike other record linkage approaches, our algorithm can also use physical models, such as body mass index curves, as prior information for record linkage. We have shown our framework can create longitudinal samples where none existed and validate pre-existing patient identifiers. We have demonstrated that in this specific case this automated approach is better than the existing identifiers., Competing Interests: The authors have declared that no competing interests exist.

In this study, an improved Data INterpolating Empirical Orthogonal Functions (DINEOF) algorithm for determination of missing values in a spatio-temporal dataset is presented. Compared with the ordinary DINEOF algorithm, the iterative reconstruction procedure until convergence based on every fixed EOF to determine the optimal EOF mode is not necessary and the convergence criterion is only reached once in the improved DINEOF algorithm. Moreover, in the ordinary DINEOF algorithm, after optimal EOF mode determination, the initial matrix with missing data will be iteratively reconstructed based on the optimal EOF mode until the reconstruction is convergent. However, the optimal EOF mode may be not the best EOF for some reconstructed matrices generated in the intermediate steps. Hence, instead of using asingle EOF to fill in the missing data, in the improved algorithm, the optimal EOFs for reconstruction are variable (because the optimal EOFs are variable, the improved algorithm is called VE-DINEOF algorithm in this study). To validate the accuracy of the VE-DINEOF algorithm, a sea surface temperature (SST) data set is reconstructed by using the DINEOF, I-DINEOF (proposed in 2015) and VE-DINEOF algorithms. Four parameters (Pearson correlation coefficient, signal-to-noise ratio, root-mean-square error, and mean absolute difference) are used as a measure of reconstructed accuracy. Compared with the DINEOF and I-DINEOF algorithms, the VE-DINEOF algorithm can significantly enhance the accuracy of reconstruction and shorten the computational time.

A conventional study design among medical and biological experimentalists involves collecting multiple measurements from a study subject. For example, experiments utilizing mouse models in neuroscience often involve collecting multiple neuron measurements per mouse to increase the number of observations without requiring a large number of mice. This leads to a form of statistical dependence referred to as clustering. Inappropriate analyses of clustered data have resulted in several recent critiques of neuroscience research that suggest the bar for statistical analyses within the field is set too low. We compare naïve analytical approaches to marginal, fixed-effect, and mixed-effect models and provide guidelines for when each of these models is most appropriate based on study design. We demonstrate the influence of clustering on a between-mouse treatment effect, a within-mouse treatment effect, and an interaction effect between the two. Our analyses demonstrate that these statistical approaches can give substantially different results, primarily when the analyses include a between-mouse treatment effect. In a novel analysis from a neuroscience perspective, we also refine the mixed-effect approach through the inclusion of an aggregate mouse-level counterpart to a within-mouse (neuron level) treatment as an additional predictor by adapting an advanced modeling technique that has been used in social science research and show that this yields more informative results. Based on these findings, we emphasize the importance of appropriate analyses of clustered data, and we aim for this work to serve as a resource for when one is deciding which approach will work best for a given study.

The lack of reproducibility with animal phenotyping experiments is a growing concern among the biomedical community. One contributing factor is the inadequate description of statistical analysis methods that prevents researchers from replicating results even when the original data are provided. Here we present PhenStat--a freely available R package that provides a variety of statistical methods for the identification of phenotypic associations. The methods have been developed for high throughput phenotyping pipelines implemented across various experimental designs with an emphasis on managing temporal variation. PhenStat is targeted to two user groups: small-scale users who wish to interact and test data from large resources and large-scale users who require an automated statistical analysis pipeline. The software provides guidance to the user for selecting appropriate analysis methods based on the dataset and is designed to allow for additions and modifications as needed. The package was tested on mouse and rat data and is used by the International Mouse Phenotyping Consortium (IMPC). By providing raw data and the version of PhenStat used, resources like the IMPC give users the ability to replicate and explore results within their own computing environment.

Background: Over recent years there has been a strong movement towards the improvement of vital statistics and other types of health data that inform evidence-based policies. Collecting such data is not cost free. To date there is no systematic framework to guide investment decisions on methods of data collection for vital statistics or health information in general. We developed a framework to systematically assess the comparative costs and outcomes/benefits of the various data methods for collecting vital statistics., Methodology: The proposed framework is four-pronged and utilises two major economic approaches to systematically assess the available data collection methods: cost-effectiveness analysis and efficiency analysis. We built a stylised example of a hypothetical low-income country to perform a simulation exercise in order to illustrate an application of the framework., Findings: Using simulated data, the results from the stylised example show that the rankings of the data collection methods are not affected by the use of either cost-effectiveness or efficiency analysis. However, the rankings are affected by how quantities are measured., Conclusion: There have been several calls for global improvements in collecting useable data, including vital statistics, from health information systems to inform public health policies. Ours is the first study that proposes a systematic framework to assist countries undertake an economic evaluation of DCMs. Despite numerous challenges, we demonstrate that a systematic assessment of outputs and costs of DCMs is not only necessary, but also feasible. The proposed framework is general enough to be easily extended to other areas of health information.