Dataset Information

A counterfactual approach to bias and effect modification in terms of response types.

ABSTRACT: BACKGROUND: The counterfactual approach provides a clear and coherent framework to think about a variety of important concepts related to causation. Meanwhile, directed acyclic graphs have been used as causal diagrams in epidemiologic research to visually summarize hypothetical relations among variables of interest, providing a clear understanding of underlying causal structures of bias and effect modification. In this study, the authors aim to further clarify the concepts of bias (confounding bias and selection bias) and effect modification in the counterfactual framework. METHODS: The authors show how theoretical data frequencies can be described by using unobservable response types both in observational studies and in randomized controlled trials. By using the descriptions of data frequencies, the authors show epidemiologic measures in terms of response types, demonstrating significant distinctions between association measures and effect measures. These descriptions also demonstrate sufficient conditions to estimate effect measures in observational studies. To illustrate the ideas, the authors show how directed acyclic graphs can be extended by integrating response types and observed variables. RESULTS: This study shows a hitherto unrecognized sufficient condition to estimate effect measures in observational studies by adjusting for confounding bias. The present findings would provide a further understanding of the assumption of conditional exchangeability, clarifying the link between the assumptions for making causal inferences in observational studies and the counterfactual approach. The extension of directed acyclic graphs using response types maintains the integrity of the original directed acyclic graphs, which allows one to understand the underlying causal structure discussed in this study. CONCLUSIONS: The present findings highlight that analytic adjustment for confounders in observational studies has consequences quite different from those of physical control in randomized controlled trials. In particular, the present findings would be of great use when demonstrating the inherent distinctions between observational studies and randomized controlled trials.

SUBMITTER: Suzuki E

PROVIDER: S-EPMC3765813 | biostudies-other | 2013

REPOSITORIES: biostudies-other

ACCESS DATA

Publications

A counterfactual approach to bias and effect modification in terms of response types.

Suzuki Etsuji E Mitsuhashi Toshiharu T Tsuda Toshihide T Yamamoto Eiji E

BMC medical research methodology 20130731

<h4>Background</h4>The counterfactual approach provides a clear and coherent framework to think about a variety of important concepts related to causation. Meanwhile, directed acyclic graphs have been used as causal diagrams in epidemiologic research to visually summarize hypothetical relations among variables of interest, providing a clear understanding of underlying causal structures of bias and effect modification. In this study, the authors aim to further clarify the concepts of bias (confou ...[more]

PMID: 23902658

Similar Datasets

Project description:We propose a BlackBox Counterfactual Explainer, designed to explain image classification models for medical applications. Classical approaches (e.g., , saliency maps) that assess feature importance do not explain how imaging features in important anatomical regions are relevant to the classification decision. Such reasoning is crucial for transparent decision-making in healthcare applications. Our framework explains the decision for a target class by gradually exaggerating the semantic effect of the class in a query image. We adopted a Generative Adversarial Network (GAN) to generate a progressive set of perturbations to a query image, such that the classification decision changes from its original class to its negation. Our proposed loss function preserves essential details (e.g., support devices) in the generated images. We used counterfactual explanations from our framework to audit a classifier trained on a chest X-ray dataset with multiple labels. Clinical evaluation of model explanations is a challenging task. We proposed clinically-relevant quantitative metrics such as cardiothoracic ratio and the score of a healthy costophrenic recess to evaluate our explanations. We used these metrics to quantify the counterfactual changes between the populations with negative and positive decisions for a diagnosis by the given classifier. We conducted a human-grounded experiment with diagnostic radiology residents to compare different styles of explanations (no explanation, saliency map, cycleGAN explanation, and our counterfactual explanation) by evaluating different aspects of explanations: (1) understandability, (2) classifier's decision justification, (3) visual quality, (d) identity preservation, and (5) overall helpfulness of an explanation to the users. Our results show that our counterfactual explanation was the only explanation method that significantly improved the users' understanding of the classifier's decision compared to the no-explanation baseline. Our metrics established a benchmark for evaluating model explanation methods in medical images. Our explanations revealed that the classifier relied on clinically relevant radiographic features for its diagnostic decisions, thus making its decision-making process more transparent to the end-user.

Project description:BACKGROUND:Automatically activated cognitive motivational processes such as the tendency to attend to or approach smoking-related stimuli (ie, attentional and approach bias) have been related to smoking behaviors. Therefore, these cognitive biases are thought to play a role in maintaining smoking behaviors. Cognitive biases can be modified with cognitive bias modification (CBM), which holds promise as an easy-access and low-cost online intervention. However, little is known about the effectiveness of online interventions combining two varieties of CBM. Targeting multiple cognitive biases may improve treatment outcomes because these biases have been shown to be relatively independent. OBJECTIVE:This study aimed to test the individual and combined effects of two web-based CBM varieties-attentional bias modification (AtBM) and approach bias modification (ApBM)-in a double-blind randomized controlled trial (RCT) with a 2 (AtBM: active versus sham) × 2 (ApBM: active versus sham) factorial design. METHODS:A total of 504 adult smokers seeking online help to quit smoking were randomly assigned to 1 of 4 experimental conditions to receive 11 fully automated CBM training sessions. To increase participants' intrinsic motivation to change their smoking behaviors, all participants first received brief, automated, tailored feedback. The primary outcome was point prevalence abstinence during the study period. Secondary outcomes included daily cigarette use and attentional and approach bias. All outcomes were repeatedly self-assessed online from baseline to the 3-month follow-up. For the examination of training effects on outcome changes, an intention-to-treat analysis with a multilevel modeling (MLM) approach was adopted. RESULTS:Only 10.7% (54/504) of the participants completed all 11 training sessions, and 8.3% (42/504) of the participants reached the 3-month follow-up assessment. MLM showed that over time, neither AtBM or ApBM nor a combination of both differed from their respective sham training in point prevalence abstinence rates (P=.17, P=.56, and P=.14, respectively), and in changes in daily cigarette use (P=.26, P=.08, and P=.13, respectively), attentional bias (P=.07, P=.81, and P=.15, respectively), and approach bias (P=.57, P=.22, and P=.40, respectively), while daily cigarette use decreased over time across conditions for all participants (P<.001). CONCLUSIONS:This RCT provides no support for the effectiveness of combining AtBM and ApBM in a self-help web-based smoking cessation intervention. However, this study had a very high dropout rate and a very low frequency of training usage, indicating an overall low acceptability of the intervention, which precludes any definite conclusion on its efficacy. We discuss how this study can inform future designs and settings of online CBM interventions. TRIAL REGISTRATION:Netherlands Trial Register NTR4678; https://www.trialregister.nl/trial/4678.

Project description:Emotions play an important role in human communication, and the daily-life interactions of young children often include situations that require the verbalization of emotional states with verbal means, e.g., with emotion terms. Through them, one can express own emotional states and those of others. Thus, the acquisition of emotion terms allows children to participate more intensively in social contexts - a basic requirement for learning new words and for elaborating socio-emotional skills. However, little is known about how children acquire and process this specific word category, which is positioned between concrete and abstract words. In particular, the influence of valence on emotion word processing during childhood has not been sufficiently investigated. Previous research points to an advantage of positive words over negative and neutral words in word processing. While previous studies found valence effects to be influenced by factors such as arousal, frequency, concreteness, and task, it is still unclear if and how valence effects are also modified by age. The present study compares the performance of children aged from 5 to 12 years and adults in two experimental tasks: lexical decision (word or pseudoword) and emotional categorization (positive or negative). Stimuli consisted of 48 German emotion terms (24 positive and 24 negative) matched for arousal, concreteness, age of acquisition, word class, word length, morphological complexity, frequency, and neighborhood density. Results from both tasks reveal two developmental trends: First, with increasing age children responded faster and more correctly, suggesting that emotion vocabulary gradually becomes more stable and differentiated during middle childhood. Second, the influence of valence varied with age: younger children (5- and 6-year-olds) showed significantly higher performance levels for positive emotion terms compared to negative emotion terms, whereas older children and adults did not. This age-related valence effect in emotion word processing will be discussed with respect to linguistic and methodological aspects.

Dataset Information

A counterfactual approach to bias and effect modification in terms of response types.

Publications

A counterfactual approach to bias and effect modification in terms of response types.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets