Bias Control Research Articles

Synthesizing Subject-matter Expertise for Variable Selection in Causal Effect Estimation: A Case Study.

Causal graphs are an important tool for covariate selection but there is limited applied research on how best to create them. Here, we used data from the Coronary Drug Project trial to assess a range of approaches to directed acyclic graph (DAG) creation. We focused on the effect of adherence on mortality in the placebo arm, since the true causal effect is believed with a high degree of certainty. We created DAGs for the effect of placebo adherence on mortality using different approaches for identifying variables and links to include or exclude. For each DAG, we identified minimal adjustment sets of covariates for estimating our causal effect of interest and applied these to analyses of the Coronary Drug Project data. When we used only baseline covariate values to estimate the cumulative effect of placebo adherence on mortality, all adjustment sets performed similarly. The specific choice of covariates had minimal effect on these (biased) point estimates, but including nonconfounding prognostic factors resulted in smaller variance estimates. When we additionally adjusted for time-varying covariates of adherence using inverse probability weighting, covariates identified from the DAG created by focusing on prognostic factors performed best. Theoretical advice on covariate selection suggests that including prognostic factors that are not exposure predictors can reduce variance without increasing bias. In contrast, for exposure predictors that are not prognostic factors, inclusion may result in less bias control. Our results empirically confirm this advice. We recommend that hand-creating DAGs begin with the identification of all potential outcome prognostic factors.

Investigation of Cu-doped amorphous carbon film in improving corrosion resistance and interfacial conductivity of proton exchange membrane fuel cell

This study investigated Cu-doped a-C films prepared under a substrate bias gradually reducing from 120 to 60 V and different Cu target currents of 0, 0.15, 0.2, 0.3, 0.5 A, respectively, using closed field unbalanced magnetron sputtering physical vapour deposition technique. While the Cu-doped a-C film at the lowest Cu target current of 0.15 A (A0.15) remained an amorphous microstructure, Cu nanoprecipitates occurred at Cu target currents of 0.2–0.5 A. All the Cu-doped a-C films prepared exhibited good combination of corrosion resistance to a corrosive chemical mixture of 0.5 M H2SO4 + 5 ppm HF (as suggested by the corrosion current densities) and interfacial electric conductivity (as demonstrated by the interfacial contact resistance, ICR). Among all coatings, A0.15 showed the best corrosion resistance and interfacial electric conductivity, which is attributable to the precipitate-free Cu-doped amorphous microstructure. The measured current density of A0.15 was 1.95 × 10−7 A/cm2 at 0.6 V (vs. SCE) during potentiodynamic polarization, which was lower than that of the Cu-free a-C film at 7.34 × 10−7 A/cm2. The ICR value of A0.15 was 4.40 mΩcm2 at an assembly pressure of 1.4 MPa, which is about one third of the Cu-free a-C film. This study demonstrates a new deposition strategy combining Cu-doping and the control of substrate bias to improve the corrosion resistance and interfacial electric conductivity of a-C films for PEMFC.

The carbon premium: Correlation or causality? Evidence from S&P 500 companies

In the context of whether investors are aware of carbon-related risks, it is often hypothesized that there may be a carbon premium in the value of stocks of firms, conferring an abnormal excess value to firms’ shares as a form of compensation to investors for their transition risk exposure through the ownership of carbon intensive stocks. However, there is little consensus in the literature regarding the existence of such a premium. Moreover few studies have examined whether the correlation that is often observed is actually causal. The pertinent question is whether more polluting firms give higher returns or do firms with high returns have less incentive to decarbonize? In this study, we investigate whether firms’ emissions is causally linked to the presence of a carbon premium in a panel of 141 firms listed in the S&P 500 index using fixed-effects analysis, with propensity score weighting to control for selection bias in which firms increase their emissions. We find that there is a statistically significant positive carbon premium associated with Scope 1 emissions, while there is no significant premium associated with Scope 2 emissions, implying that risks associated with direct emissions by the firm are priced, while bought emissions are not.

Negative attention bias and attentional control as mechanisms in the association between insomnia and depression in young people

Evidence supports a causal role of insomnia in the development and maintenance of depression, yet mechanisms underlying this association in young people are not well established. Attention biases have been implicated separately in the sleep and depression fields and represents an important candidate mechanism. Poor sleep may lead to a negative attention bias (characteristic of depression) by impacting attentional control. This study assessed the hypothesis that attentional control and negative attention bias would sequentially mediate the relationship between insomnia and depressive symptoms in an unselected sample of young people (17–24 years). Concerns have been raised regarding the psychometric properties of tasks used to measure attention bias, and a Dual-Probe Task is emerging as a more reliable measure. Participants (N = 275, Male = 59, Mage = 19.40) completed the Dual-Probe Task, a behavioural measure of attentional control, and self-report measures of insomnia and depression. Participants completed a one-week sleep diary. Results were consistent with negative attention bias, but not attentional control, as a mechanism which partially accounts for the relationship between sleep (i.e., insomnia severity, sleep duration, sleep efficiency, sleep latency) and depression. This study highlights sleep and negative attention bias as potentially modifiable risk factors to reduce depressive symptoms in young people.

Attention deficit hyperactivity disorder and dental traumatic injuries: a systemic review and meta-analysis.

The objective of this study was to assess whether individuals with attention deficit hyperactivity disorder (ADHD) exhibit a higher prevalence of traumatic dental injuries (TDIs) compared to those without ADHD through a systematic review and meta-analysis of the existing literature. A search strategy using the Medical Subject Heading (MeSH) vocabulary was employed for a comprehensive search across various databases, including PubMed, Scopus, Web of Science, Cochrane Library, and Embase. The Joanna Briggs Institute Summary was utilized for data collection. Additionally, quality assessment, meta-analysis, and bias control were conducted to ensure the reliability of the included studies. A meta-analysis was performed to consolidate the findings of the individual studies. The prevalence of TDIs among individuals with ADHD ranged from 9.6% to 68.2%, while in the healthy control group, it ranged from 0.8% to 44.7%. The meta-analysis findings revealed that individuals with ADHD had 1.98 times higher odds (OR = 1.98, ranging from 1.51 to 2.59 with 95% CI) of experiencing TDIs compared to individuals without ADHD. The findings of this study suggest a significant association between ADHD and an increased risk of TDIs. Individuals with ADHD were found to be nearly twice as likely to experience TDIs compared to those without ADHD. Efforts should not only be directed towards improving the oral health of this vulnerable group of individuals, but also healthcare practitioners need to be provided with opportunities to create awareness and implement preventive measures to mitigate the risk of TDIs among individuals with ADHD.

A configurable ultra-low noise current source for transition-edge sensor characterization

Transition-edge sensors (TESs) are sensitive devices for detecting photons from millimeter radiation to γ rays. Their photon counting efficiency and collecting area benefit from large-array multiplexing scheme, and therefore the development of multiplexing readout system has been an important topic in this field. Among the many multiplex techniques, time-division multiplexing (TDM) superconducting quantum interference device (SQUID) has been used most widely for TES readout. In this work, we design a Configurable Ultra-Low Noise Current Source (CLCS) for TES characterization and as a part of a whole TDM-TES bias control system. The CLCS is based on the feedback structure of ultra-low noise instrumentation amplifiers and low-noise, high-resolution (20 bits) digital-to-analog converter (DAC). CLCS has an ultra-high resolution of 10 nA in the 0 to 5 mA current output range, and can perform current-voltage (IV) sweep and bias-step tests to measure key TES parameters on board. The feedback structure of the CLCS also avoids the issue of impedance mismatch.

Dynamic Environment Responsive Online Meta-Learning with Fairness Awareness

The fairness-aware online learning framework has emerged as a potent tool within the context of continuous lifelong learning. In this scenario, the learner’s objective is to progressively acquire new tasks as they arrive over time, while also guaranteeing statistical parity among various protected sub-populations, such as race and gender when it comes to the newly introduced tasks. A significant limitation of current approaches lies in their heavy reliance on the i.i.d (independent and identically distributed) assumption concerning data, leading to a static regret analysis of the framework. Nevertheless, it’s crucial to note that achieving low static regret does not necessarily translate to strong performance in dynamic environments characterized by tasks sampled from diverse distributions. In this article, to tackle the fairness-aware online learning challenge in evolving settings, we introduce a unique regret measure, FairSAR, by incorporating long-term fairness constraints into a strongly adapted loss regret framework. Moreover, to determine an optimal model parameter at each time step, we introduce an innovative adaptive fairness-aware online meta-learning algorithm, referred to as FairSAOML. This algorithm possesses the ability to adjust to dynamic environments by effectively managing bias control and model accuracy. The problem is framed as a bi-level convex-concave optimization, considering both the model’s primal and dual parameters, which pertain to its accuracy and fairness attributes, respectively. Theoretical analysis yields sub-linear upper bounds for both loss regret and the cumulative violation of fairness constraints. Our experimental evaluation of various real-world datasets in dynamic environments demonstrates that our proposed FairSAOML algorithm consistently outperforms alternative approaches rooted in the most advanced prior online learning methods.

Design of clinical trials in integrative medicine: The issue of personalization

IntroductionClassic randomized controlled trial design provides information about the effectiveness of a treatment at population level. In clinical practice based on precision medicine, we seek study designs that provide estimates accounting for individual patients’ demographics and context. Integrative medicine has a strong emphasis on personalization, and there is a pressing need for advancements in trial design to support clinical decisions. DiscussionTo enhance the generalizability and implementation of research evidence, study designs first need to mimic the real-world from both the clinicians’ and patients’ perspectives, which could be optimized based on stakeholder analysis. Previous stakeholder analyzes showed that the lack of personalization in clinical trial design is a key challenge for the adoption of evidence clinically. Variations of randomized controlled trial design including better-powered and better-designed subgroup analysis, pragmatic design, the N-of-1 cross-over design, and adaptive design have been introduced to personalize clinical trials. Each of these variations has unique advantages, assumptions and limitations. Stratified randomization can be used to divide a disease population into subgroups according to different schools of theory to match with the corresponding treatment, while preserving the generalizability to general readership. Patient-level meta-analysis can improve the power and precision of subgroup analysis. Enrichment design can increase the efficiency of a trial in detecting treatment effect(s) and in mimicking actual clinical use. Practical issues, including bias control and assessment in pragmatic trials, and N-of-1 trials are also discussed with recent trials as examples. We suggest strategies to maximize the validity and translation potential of clinical trials in integrative medicine with pragmatic or personalized design. ConclusionVariations of randomized controlled trial design improve the precision of estimates at patient level, which should be carefully considered in the trial design of integrative medicine according to the context of the future application of evidence.

Fabrication of thin-film batteries composed of LiCoO2, Li3PO4, and Li layers

AbstractThis paper reports the fabrication of thin-film batteries which are composed of three stacking layers: LiCoO2, Li3PO4, and Li. First, a LiCoO2 layer is constructed on an electron-conductive substrate by pulsed laser deposition as a cathode. The crystallinity of the LiCoO2 layer is mainly controlled by the cationic ratio of Li and Co. Subsequently, an amorphous Li3PO4 layer with a high ionic conductivity is further deposited on the cathode LiCoO2 layer by radio frequency magnetron sputtering as a solid electrolyte. To avoid any possible damage which causes the formation of resistive species between LiCoO2 and Li3PO4, bias control of the substrate during Li3PO4 deposition is essential. Finally, a Li metal layer is deposited as an anode/current collector on the Li3PO4/LiCoO2 bilayer by resistive heating evaporation in a vacuum at an elevated temperature for the formation of a low resistive interface. The fabricated three-layer thin-film battery shows a high-rate capability when the LiCoO2 layer is a (104)-oriented epitaxial film.

Attentional Bias Toward Cigarette-Related Cues in Male Smokers is Modulated by Cognitive Control Mechanisms

Background: Many studies have found that smokers’ attentional bias toward cigarette-related cues and cognitive control impairment significantly impacts their cigarette use. However, there is limited research on how the interaction between attentional bias and cognitive control may modulate smokers’ cigarette-seeking behavior. Objectives: This study used a cigarette Stroop task to examine whether smokers with different attentional control ability had different levels of attentional bias toward cigarette-related cues. Methods: A total of 130 male smokers completed the Flanker task to measure their attentional control ability. The attentional control scores of all participants were ranked from low to high, with the top 27% placed in the high attentional control group and the bottom 27% in the low attentional control group. Subsequently, both groups completed the cigarette Stroop task to measure their attentional bias toward cigarette-related cues. Results: Smokers with low attentional control responded more slowly to cigarette-related cues than to neutral cues, while smokers with high attentional control showed no significant difference in their response time to either condition. Conclusions/Importance: Attentional control ability can regulate smokers’ attentional bias toward cigarette-related cues. Smokers with low attentional control ability are more likely to have attentional bias toward cigarette-related cues, offering insights for targeted prevention of cigarette addiction.

Do Implant Coatings Affect Healing of Placed Implants? An Umbrella Review.

To analyze the available evidence and assess the effect of different implant coatings on healing outcomes. Using the PICOS strategy, a structured question was formed. A protocol was agreed upon and registered with PROSPERO (no. CRD42022321926). The MEDLINE, Embase, Cochrane Database of Systematic Reviews, Scopus, Web of Science, Pubmed, and ScienceDirect databases were searched using a structured strategy. Study selection was independently carried out in duplicate, first by title and abstract, then by full-text assessment. Quality and risk of bias were independently assessed in duplicate using AMSTAR 2 and ROBIS. Data extraction was independently undertaken in duplicate using a predefined extraction form. The search yielded 11 systematic reviews for inclusion. The most commonly assessed coatings were based on calcium phosphate-including hydroxyapatite (HA), brushite, and bioabsorbable nano-HA-followed by bisphosphonate, then bioactive glass coatings. Included reviews most frequently assessed marginal bone loss (MBL), bone-to-implant contact (BIC), and survival/success rates. There was considerable heterogeneity and small sample sizes. The quality assessment suggested low confidence in the reviews and high risk of bias. The included reviews provide weak evidence that implant coatings improve osseointegration and reduce MBL following implant placement. There was weak evidence for progressive complications for calcium phosphate coatings. Further research and long-term multicenter controlled clinical trials with improved standardization and control of bias are required to better understand the effects of coating implants.

Does digital transformation promote innovation performance? Evidence from listed Chinese firms

This study examines the impact of digitalization on firm innovation. Using a comprehensive sample of publicly listed Chinese firms, we construct a micro-level digitalization indicator using a textual analysis of annual financial reports. It was found that digitalization leads to a significant upswing in a firm’s innovative output. Rigorous robustness checks and stability tests were conducted, including variable substitutions and different models, such as the Tobit and Poisson models. Additionally, the study utilizes a treatment effects model to account for unobserved heterogeneity and control for potential biases, ensuring the reliability and validity of the findings. Whether an enterprise is state-owned (SOEs) or non-state owned (non-SOEs), the digitalization of enterprises has dramatically enhanced their innovation capabilities. Digitalization has a significant effect on financially constrained firms. Compared to non-SOEs, the effect of financial constraints is more pronounced among SOEs. Highly digitalized firms also experienced higher growth rates and lower leverage ratios relative to firms with low digitalization, and they are more likely to receive governmental subsidies. Furthermore, the findings suggest that policymakers should prioritize initiatives aimed at promoting digitalization across all sectors of the economy. By fostering an environment conducive to digital innovation, governments can stimulate economic growth and enhance firms' competitiveness on a global scale. In essence, the study underscores the transformative potential of digitalization in driving innovation and economic progress.

Towards a Clearer Causal Question Underlying the Association Between Cancer and Dementia.

Several observational studies have described an inverse association between cancer diagnosis and subsequent dementia risk. Multiple biologic mechanisms and potential biases have been proposed in attempts to explain this association. One proposed explanation is the opposite expression of Pin1 in cancer and dementia, and we use this explanation and potential drug target to illustrate the required assumptions and potential sources of bias for inferring an effect of Pin1 on dementia risk from analyses measuring cancer diagnosis as a proxy for Pin1 expression. We used data from the Rotterdam Study, a population-based cohort. We estimate the association between cancer diagnosis (as a proxy for Pin1) and subsequent dementia diagnosis using two different proxy methods and with confounding and censoring for death addressed with inverse probability weights. We estimate and compare the complements of a weighted Kaplan-Meier survival estimator at 20 years of follow-up. Out of 3634 participants, 899 (25%) were diagnosed with cancer, of whom 53 (6%) had dementia, and 567 (63%) died. Among those without cancer, 15% (411) were diagnosed with dementia, and 667 (24%) died over follow-up. Depending on the confounding and selection bias control, and the way in which cancer was used as a time-varying proxy exposure, the risk ratio for dementia diagnosis ranged from 0.71 (95% confidence interval [CI] = 0.49, 0.95) to 1.1 (95% CI = 0.79, 1.3). Being explicit about the underlying mechanism of interest is key to maximizing what we can learn from this cancer-dementia association given available or readily collected data, and to defining, detecting, and preventing potential biases.

Fatal Flaws in Voice Research and How to Avoid Them, Part Two: Qualitative Studies

Abstract: This article is the second part of a two-part series examining so-called “fatal flaws” in voice science, with an emphasis on qualitative studies. This type of research looks for evidence to explore real-world phenomena, and asks questions like, “why?” or “I wonder what, or how?” The most common errors noted in qualitative studies include researchers’ overestimation of their expertise, failure to practice intellectual humility, failure to control for researcher bias, and privileging the “expert in the room” rather than participants’ perspectives. Interview protocols are examined, including a common method for ensuring trustworthiness called “member checking.” Recommendations for avoiding common errors in survey research, both qualitative and quantitative, are discussed.

Controlling underestimation bias in reinforcement learning via minmax operation

Obtaining the accurate value estimation and reducing the estimation bias are the key issues in reinforcement learning. However, current methods that address the overestimation problem tend to introduce underestimation, which face a challenge of precise decision-making in many fields. To address this issue, we conduct a theoretical analysis of the underestimation bias and propose the minmax operation, which allow for flexible control of the estimation bias. Specifically, we select the maximum value of each action from multiple parallel state-action networks to create a new state-action value sequence. Then, a minimum value is selected to obtain more accurate value estimations. Moreover, based on the minmax operation, we propose two novel algorithms by combining Deep Q-Network (DQN) and Double DQN (DDQN), named minmax-DQN and minmax-DDQN. Meanwhile, we conduct theoretical analyses of the estimation bias and variance caused by our proposed minmax operation, which show that this operation significantly improves both underestimation and overestimation biases and leads to the unbiased estimation. Furthermore, the variance is also reduced, which is helpful to improve the network training stability. Finally, we conduct numerous comparative experiments in various environments, which empirically demonstrate the superiority of our method.

Deep reinforcement learning enables better bias control in benchmark for virtual screening

Virtual screening (VS) has been incorporated into the paradigm of modern drug discovery. This field is now undergoing a new wave of revolution driven by artificial intelligence and more specifically, machine learning (ML). In terms of those out-of-the-box datasets for model training or benchmarking, their data volume and applicability domain are limited. They are suffering from the biases constantly reported in the ML application. To address these issues, we present a novel benchmark named MUBDsyn. The utilization of synthetic decoys (i.e., presumed inactives) is the main feature of MUBDsyn, where deep reinforcement learning was leveraged for bias control during decoy generation. Then, we carried out extensive validations on this new benchmark. First, we confirmed that MUBDsyn was superior to the classical benchmarks in control of domain bias, artificial enrichment bias and analogue bias. Moreover, we found that the assessment of ML models based on MUBDsyn was less biased as revealed by the analysis of asymmetric validation embedding bias. In addition, MUBDsyn showed better setting of benchmarking challenge for deep learning models compared with NRLiSt-BDB. Overall, we have proven that MUBDsyn is the close-to-ideal benchmark for VS. The computational tool is publicly available for the easy extension of MUBDsyn.

Degrees of vulnerability to poverty: a low-income dynamics approach for Chile

AbstractI propose an empirical framework to identify different degrees of vulnerability to poverty using two vulnerability lines that classify currently non-poor people into risk groups: high, moderate and low risk of falling into poverty in the next period. The latter corresponds to the income secure middle class. My approach makes two contributions. First, it extends recent research that defines the middle class using a vulnerability threshold by introducing a new subdivision of the vulnerable group that would be useful in practice for public policy objectives. Second, it uses two models to predict both the probability of entering poverty and household income as part of the estimation procedures. The former controls for initial conditions effects and attrition bias, and the latter addresses the retransformation problem. I apply my approach to Chile using longitudinal data from the P-CASEN 2006–2009. The resulting vulnerability cut-offs (using the upper-middle-income country poverty line) are $20.0 per person per day for the low vulnerability line and $9.9 pppd for the high vulnerability line (both in 2011 PPP). My vulnerability lines differ significantly from those estimated in previous research on vulnerability and the middle class in Latin America. I argue that previous research has underestimated the size of the population at risk of falling into poverty and overestimated the growth of the middle class. Misclassifying the vulnerable as middle class limits their access to anti-poverty policies.

Impact of social desirability bias on the CoPsoQ-ISTAS 21 psychosocial risk assessment questionnaire.

Self-reported test is one of the main psychosocial risk assessment tools. However, this test it is susceptible to certain sources of error, including social desirability. Since psychosocial risks are emerging, there are not many studies on their assessment. The aim of this work is to analyze the impact of social desirability on the short version of the CopSoq-ISTAS 21 assessment tool. A total of 563 workers (45.10% women and 54.90% men) participated in this study. The short version of the CoPsoQ-Istas21 questionnaire with four Likert scale questions as markers, which correspond to the Eysenck Personality Lie Scale Questionnaire Revised (EPQ-r), were used. The sample was divided into two halves, and both a confirmatory analysis and an exploratory analysis were carried out to find out the factorial structure of the scale and, with it, apply the bias filtering method. The results indicate that 10% of the scale is biased due to social desirability, and that there are significant differences between the group with bias clean scores and the group with scores without bias control. The effects of social desirability on the scale are verified, so it is concluded that in a psychosocial risk assessment is not enough to apply a self-report test and interpret its results, being necessary to minimize the sources of error.

What are we really decoding? Unveiling biases in EEG-based decoding of the spatial focus of auditory attention

Objective. Spatial auditory attention decoding (Sp-AAD) refers to the task of identifying the direction of the speaker to which a person is attending in a multi-talker setting, based on the listener’s neural recordings, e.g. electroencephalography (EEG). The goal of this study is to thoroughly investigate potential biases when training such Sp-AAD decoders on EEG data, particularly eye-gaze biases and latent trial-dependent confounds, which may result in Sp-AAD models that decode eye-gaze or trial-specific fingerprints rather than spatial auditory attention. Approach. We designed a two-speaker audiovisual Sp-AAD protocol in which the spatial auditory and visual attention were enforced to be either congruent or incongruent, and we recorded EEG data from sixteen participants undergoing several trials recorded at distinct timepoints. We trained a simple linear model for Sp-AAD based on common spatial patterns filters in combination with either linear discriminant analysis (LDA) or k-means clustering, and evaluated them both across- and within-trial. Main results. We found that even a simple linear Sp-AAD model is susceptible to overfitting to confounding signal patterns such as eye-gaze and trial fingerprints (e.g. due to feature shifts across trials), resulting in artificially high decoding accuracies. Furthermore, we found that changes in the EEG signal statistics across trials deteriorate the trial generalization of the classifier, even when the latter is retrained on the test trial with an unsupervised algorithm. Significance. Collectively, our findings confirm that there exist subtle biases and confounds that can strongly interfere with the decoding of spatial auditory attention from EEG. It is expected that more complicated non-linear models based on deep neural networks, which are often used for Sp-AAD, are even more vulnerable to such biases. Future work should perform experiments and model evaluations that avoid and/or control for such biases in Sp-AAD tasks.

Numerical Analysis of Dither Induced Penalty for Automatic Bias Control Methods

Numerical Analysis of Dither Induced Penalty for Automatic Bias Control Methods