Anomaly and Novelty detection for robust semi-supervised learning

Statistics and Computing - Three important issues are often encountered in Supervised and Semi-Supervised Classification: class memberships are unreliable for some training units (label noise), a...     

Adult children of parents with mental illness: navigating stigma

The experiences of children who live with parental mental illness are becoming increasingly recognized. However, there remains a limited body of knowledge in relation to an individual's longer term experiences. This study sought adult children's experiences of childhood parental mental illness. It generated reflections of 13 adult children who had lived with parents with mental illness during their childhood. The narrative design of the study facilitated a voice for participants. The paper presents one main theme of the findings. The findings offer a unique insight into childhood awareness of social stigmas and children's behavioural changes to avoid disclosure of parental mental illness. Participants noted that they were aware of social stigmas associated with mental illness during their childhood. This contributed to their fear of disclosure to others that fuelled a culture of familial secrecy, reinforcing the children's own sense of difference. Further work is required to enhance community understanding about the familial journeys of parental mental illness and the impact of negative‐natured stigmas.     

A community of singers: employing exploratory cluster analysis to reveal profiles of psychosocial characteristics among members of a street choir for homeless and formerly homeless individuals

This exploratory study employed cluster analysis to identify profiles of resilience, social support, stress, and mental health among members of a homeless and formerly homeless street choir; we examined the association between choir attendance and cluster membership, and the interaction of cluster membership and race. Respondents (N?=?111, 66% African American, 84% homeless) had participated in the choir for an average of 12 months (SD?=?12.02). We stratified the analysis according to homelessness status. Among those housed, k-means cluster analysis revealed two profiles: (1) relatively resourced but lower resilience, and (2) high resilience. Three profiles emerged among the homeless: (1) psychologically vulnerable, high stress and mental health concerns; (2) relatively resourced, showing average levels of mental health concerns and social support, minimally elevated stress, and higher resilience; and (3) low resilience. Examining the interaction, non-African American, housed participants in the high resilience cluster reported the most frequent choir attendance (M?=?4.00, SD?=?0.00); among the homeless, non-African Americans in the psychologically vulnerable cluster reported the most frequent attendance (M?=?3.88, SD?=?1.25). Results support the need for programs that address higher order needs concurrently with housing; they also highlight the ongoing need to provide particular outreach to minority populations.     

PERFORMANCE GUARANTEES FOR INDIVIDUALIZED TREATMENT RULES

Because many illnesses show heterogeneous response to treatment, there is increasing interest in individualizing treatment to patients [11]. An individualized treatment rule is a decision rule that recommends treatment according to patient characteristics. We consider the use of clinical trial data in the construction of an individualized treatment rule leading to highest mean response. This is a difficult computational problem because the objective function is the expectation of a weighted indicator function that is non-concave in the parameters. Furthermore there are frequently many pretreatment variables that may or may not be useful in constructing an optimal individualized treatment rule yet cost and interpretability considerations imply that only a few variables should be used by the individualized treatment rule. To address these challenges we consider estimation based on l(1) penalized least squares. This approach is justified via a finite sample upper bound on the difference between the mean response due to the estimated individualized treatment rule and the mean response due to the optimal individualized treatment rule.     

Model‐based clustering of longitudinal data

A new family of mixture models for the model‐based clustering of longitudinal data is introduced. The covariance structures of eight members of this new family of models are given and the associated maximum likelihood estimates for the parameters are derived via expectation–maximization (EM) algorithms. The Bayesian information criterion is used for model selection and a convergence criterion based on the Aitken acceleration is used to determine the convergence of these EM algorithms. This new family of models is applied to yeast sporulation time course data, where the models give good clustering performance. Further constraints are then imposed on the decomposition to allow a deeper investigation of the correlation structure of the yeast data. These constraints greatly extend this new family of models, with the addition of many parsimonious models. The Canadian Journal of Statistics 38:153–168; 2010 © 2010 Statistical Society of Canada     

Adaptive Confidence Intervals for the Test Error in Classification

The estimated test error of a learned classifier is the most commonly reported measure of classifier performance. However, constructing a high quality point estimator of the test error has proved to be very difficult. Furthermore, common interval estimators (e.g. confidence intervals) are based on the point estimator of the test error and thus inherit all the difficulties associated with the point estimation problem. As a result, these confidence intervals do not reliably deliver nominal coverage. In contrast we directly construct the confidence interval by use of smooth data-dependent upper and lower bounds on the test error. We prove that for linear classifiers, the proposed confidence interval automatically adapts to the non-smoothness of the test error, is consistent under fixed and local alternatives, and does not require that the Bayes classifier be linear. Moreover, the method provides nominal coverage on a suite of test problems using a range of classification algorithms and sample sizes.     

Family-Level Continuities in Childbearing in Low-Fertility Societies

A number of studies of populations in earlier generations haveshown that fertility patterns of parents and children arepositively correlated, although the relationship is frequentlydesignated as `weak'. Models that may be used to investigate theways in which patterns of demographic behaviour persist betweengenerations are considered. The principal frameworks used arefitting of simulation and multi-level models. The data sourcesutilised are the 1986 ISSP co-ordinated series of surveys onsocial networks, the country files for Italy, Norway and Polandfrom the UNECE co-ordinated FFS programme, and the US NationalSurvey of Families and Households which contains particularlyrich information on the experience of demographic events acrossdifferent generations. We find that the relationship betweenfertility of successive generations is becoming stronger withtime, and that it is now of a comparable order of magnitude towidely-used conventional covariates such as educational level. This intergenerational relationship cannot be explained bydifferential fertility across socio-economic groups. Reasons whythe strength of the relationship has been understated and theimplications of results from such analyses are discussed.     

When data justice and environmental justice meet: formulating a response to extractive logic through environmental data justice

ABSTRACT

Environmental data justice (EDJ) emerges from conversations between data justice and environmental justice while identifying the limits and tensions of these lenses. Through a reflexive process of querying our entanglement in non-innocent relations, this paper develops and engages EDJ by examining how it informs the work of the Environmental Data & Governance Initiative (EDGI), a distributed, consensus-based organization that formed in response to the 2016 US presidential election. Through grassroots archiving of data sets, monitoring federal environmental and energy agency websites, and writing rapid-response reports about how federal agencies are being undermined, EDGI mobilizes EDJ to challenge the ‘extractive logic’ of current federal environmental policy and data infrastructures. ‘Extractive logic’ disconnects data from provenance, privileges the matrix of domination, and whitewashes data to generate uncertainty. We use the dynamic EDJ framework to reflect on EDGI’s public comment advising against the US Environmental Protection Agency’s proposed rule for Transparent Science. Through EDJ, EDGI aspires to create new environmental data infrastructures and practices that are participatory and embody equitable, transparent data care.     

Estimation and Inference in Two-Step Econometric Models

A commonly used procedure in a wide class of empirical applications is to impute unobserved regressors, such as expectations, from an auxiliary econometric model. This two-step (T-S) procedure fails to account for the fact that imputed regressors are measured with sampling error, so hypothesis tests based on the estimated covariance matrix of the second-step estimator are biased, even in large samples. We present a simple yet general method of calculating asymptotically correct standard errors in T-S models. The procedure may be applied even when joint estimation methods, such as full information maximum likelihood, are inappropriate or computationally infeasible. We present two examples from recent empirical literature in which these corrections have a major impact on hypothesis testing.     

An Opportunistic Field Experiment in Community Water Conservation

This quasi-experimental, seven-year study evaluated the effectiveness of an educative versus a price structuring approach to the strategic management of domestic water consumption. The Knowledge, Attitudes, and Behavior scales of the Water Survey Questionnaire (Watson, Moore, McLachlan, Bradley & Murphy, 1988) measured the effects of the two strategies on water conservation. Two thousand six hundred parents, teachers, high, and elementary students in cross-sectional and longitudinal samples were measured in three data gathering rounds at the first, fourth, and seventh years. In the educative phase between the first and second rounds, water conservation increased. Although there was no significant change in behavior, some decline in positive attitudes and intentions occurred during the price structuring phase between the second and third rounds. The implications of these findings are discussed in relation to other studies. Findings on the stability of inter-group and inter-variable relationships and on the consistency of the measuring instrument across time are also presented and discussed.