Multiple-bias modelling for analysis of observational data

Summary.  Conventional analytic results do not reflect any source of uncertainty other than random error, and as a result readers must rely on informal judgments regarding the effect of possible biases. When standard errors are small these judgments often fail to capture sources of uncertainty and their interactions adequately. Multiple-bias models provide alternatives that allow one systematically to integrate major sources of uncertainty, and thus to provide better input to research planning and policy analysis. Typically, the bias parameters in the model are not identified by the analysis data and so the results depend completely on priors for those parameters. A Bayesian analysis is then natural, but several alternatives based on sensitivity analysis have appeared in the risk assessment and epidemiologic literature. Under some circumstances these methods approximate a Bayesian analysis and can be modified to do so even better. These points are illustrated with a pooled analysis of case–control studies of residential magnetic field exposure and childhood leukaemia, which highlights the diminishing value of conventional studies conducted after the early 1990s. It is argued that multiple-bias modelling should become part of the core training of anyone who will be entrusted with the analysis of observational data, and should become standard procedure when random error is not the only important source of uncertainty (as in meta-analysis and pooled analysis).     

Reducing Bias and Mean Squared Error Associated With Regression-Based Odds Ratio Estimators

Ratio estimators of effect are ordinarily obtained by exponentiating maximum-likelihood estimators (MLEs) of log-linear or logistic regression coefficients. These estimators can display marked positive finite-sample bias, however. We propose a simple correction that removes a substantial portion of the bias due to exponentiation. By combining this correction with bias correction on the log scale, we demonstrate that one achieves complete removal of second-order bias in odds ratio estimators in important special cases. We show how this approach extends to address bias in odds or risk ratio estimators in many common regression settings. We also propose a class of estimators that provide reduced mean bias and squared error, while allowing the investigator to control the risk of underestimating the true ratio parameter. We present simulation studies in which the proposed estimators are shown to exhibit considerable reduction in bias, variance, and mean squared error compared to MLEs. Bootstrapping provides further improvement, including narrower confidence intervals without sacrificing coverage.     

Editorial Collaborators

Abstract

The present note explores sources of misplaced criticisms of P-values, such as conflicting definitions of “significance levels” and “P-values” in authoritative sources, and the consequent misinterpretation of P-values as error probabilities. It then discusses several properties of P-values that have been presented as fatal flaws: That P-values exhibit extreme variation across samples (and thus are “unreliable”), confound effect size with sample size, are sensitive to sample size, and depend on investigator sampling intentions. These properties are often criticized from a likelihood or Bayesian framework, yet they are exactly the properties P-values should exhibit when they are constructed and interpreted correctly within their originating framework. Other common criticisms are that P-values force users to focus on irrelevant hypotheses and overstate evidence against those hypotheses. These problems are not however properties of P-values but are faults of researchers who focus on null hypotheses and overstate evidence based on misperceptions that p?=?0.05 represents enough evidence to reject hypotheses. Those problems are easily seen without use of Bayesian concepts by translating the observed P-value p into the Shannon information (S-value or surprisal) –log₂(p).     

Pathological Gambling: Influence of Quality of Life and Psychological Distress on Abstinence After Cognitive-Behavioral Inpatient Treatment

The aim of this study is to understand the influence of psychological distress and quality of life on the maintenance of therapy success during a 1 year follow-up in pathological gamblers after inpatient cognitive-behavioral treatment. In a sample of 281 pathological gamblers (247 men, 34 women) the following measures were taken: psychological distress (beginning and end of treatment, and follow up), quality of life and abstinence from gambling. The results showed that relapsed pathological gamblers suffer of higher psychological distress at discharge and of lower quality of life during follow-up than abstinent gamblers. In a mediator analysis the causal relations between the three variables could be clarified: psychological distress has an impact on abstinence during follow-up, and quality of life can be considered a mediator variable for the relation between psychological distress and abstinence from pathological gambling.     

Conformity to implicit social pressure: the role of political identity

Although studies have shown that implicit social cues, such as images of watchful eyes, can elicit prosocial behavior, little research to date has examined individual differences in people’s susceptibility to such subtle social cues. For example, individuals with a conservative ideology typically value social conformity, obedience, and adherence to social norms more than liberals. To examine partisan heterogeneity, we analyze data from two large randomized field experiments on voting behavior. Results suggest that the impact of eyespots on voter mobilization is indeed likely driven by political identity, with a significant effect for Republicans but not Independents or Democrats. These findings are consistent with an emerging line of research revealing individual differences in how susceptible humans are to implicit social cues.     

The development of infants’ sensitivity to native versus non-native rhythm

Speech rhythm is considered one of the first windows into the native language, and the taxonomy of rhythm classes is commonly used to explain early language discrimination. Relying on formal rhythm classification is problematic for two reasons. First, it is not known to which extent infants’ sensitivity to language variation is attributable to rhythm alone, and second, it is not known how infants discriminate languages not classified in any of the putative rhythm classes. Employing a central-fixation preference paradigm with natural stimuli, this study tested whether infants differentially attend to native versus nonnative varieties that differ only in temporal rhythm cues, and both of which are rhythmically unclassified. An analysis of total looking time did not detect any rhythm preferences at any age. First-look duration, arguably more closely reflecting infants’ underlying perceptual sensitivities, indicated age-specific preferences for native versus non-native rhythm: 4-month-olds seemed to prefer the native-, and 6-month-olds the non-native language-variety. These findings suggest that infants indeed acquire native rhythm cues rather early, by the 4th month, supporting the theory that rhythm can bootstrap further language development. Our data on infants’ processing of rhythmically unclassified languages suggest that formal rhythm classification does not determine infants’ ability to discriminate language varieties.     

Managing flexible work arrangements: Teleworking and output controls

Flexible work arrangements present managers with challenges regarding how to manage employees using these arrangements. To date, little research has investigated how managers address these challenges. We investigate the relationship between the use of a specific implementation of flexible work (teleworking) and control system design, specifically the emphasis on output controls. Teleworking reduces the feasibility of monitoring employee behaviour as a control mechanism. Control theory suggests that this might be compensated by placing more emphasis on output controls. We conduct a survey (N?=?897) among employees of a financial services institution, of whom 69% is allowed to telework. We find that among teleworking employees, the share of teleworking hours is positively related to the emphasis on output controls. However, employees who are allowed to telework report less emphasis on output controls by their manager relative to those not allowed to telework. We pose various directions for future research, which may help in explaining these findings.     

A macro‐financial analysis of pension system reforms in emerging Europe: The performance of IRAs and policy lessons for Serbia

The article explores the initial macro‐financial performance of partial pension system “privatizations”— involving privately‐managed individual retirement savings accounts (IRAs) — undertaken in many emerging European countries. Using empirical data for a period of close to a decade, the evidence shows that returns on privately‐managed IRAs have been below the implicit rate of return of public pay‐as‐you‐go (PAYG) systems. High operating costs and undeveloped capital markets are identified as major contributing factors to the failure of privately‐managed IRAs to meet reform expectations. In light of empirical evidence, Serbia is advised to focus on parametric PAYG reforms and to avoid reforms that involve the partial “privatization” of the pension system.