A conceptual introduction to Bayesian Model Averaging

Hinne, Max; Gronau, Quentin Frederik; Bergh, Don van den; Wagenmakers, Eric‐Jan

doi:10.31234/osf.io/wgb64

Cited by 25 publications

(31 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We further used the Bayesian model averaging (BMA) method (Hinne, Gronau, van den Bergh, & Wagenmakers, 2020 ) to weigh evidence for the interaction between cue condition and task load. This method weighs evidence for a particular effect across models that include the effect of interest against models that are stripped of the effect of interest.…”

Section: Methodsmentioning

confidence: 99%

Selection in working memory is resource-demanding: Concurrent task effects on the retro-cue effect

Lin

Sasin

Fougnie

2021

Atten Percept Psychophys

View full text Add to dashboard Cite

In a retro-cue paradigm, after memorizing a set of objects, people are cued to remember only a subset. Improved memory from the retro-cue suggests that selection processes can benefit items stored in working memory. Does selection in working memory require attention? If so, an attention-demanding task should disrupt retro-cue effects. Studies using a dual-task paradigm have found mixed results, with only one study (Janczyk & Berryhill, Attention, Perception, and Psychophysics, 76 (3), 715–724, 2014) showing a decreased retro-cue effect by a secondary task. Here we explore a potential issue in that study – the temporal overlap of the secondary task response with the memory test presentation. This raises questions about whether the secondary task was impairing selection processes in memory or was impacting the memory response. We replicated their paradigm by inserting a tone discrimination task at the retro-cue offset, but we also included a condition in which the tone task and the memory test were temporally separated. In Experiment 1, performing the tone task did not impair the retro-cue effect. In Experiment 2, we added an articulatory suppression task as in Janczyk and Berryhill’s study, and we found that the requirement to execute the tone task impaired retro-cue effects. This impairment was independent of whether the tone and memory tasks overlapped. These findings suggest that internal prioritization can be impaired by dual-task interference, but may only occur when such interference is robust enough, for example, due to switching between multiple tasks.

show abstract

Section: Methodsmentioning

confidence: 99%

Selection in working memory is resource-demanding: Concurrent task effects on the retro-cue effect

Lin

Sasin

Fougnie

2021

Atten Percept Psychophys

View full text Add to dashboard Cite

show abstract

“…Bayesian Model Averaging is closely related to linear pooling methods because both methods provide a linear combination of the elicitation results. However, while the weights can be determined independent of the data in linear pooling, they are defined as the Bayesian posterior model probability in Bayesian model averaging (Hinne et al, 2019). Bayesian…”

Section: Mathematical Aggregationmentioning

confidence: 99%

Practical challenges and methodological flexibility in prior elicitation.

Stefan¹,

Evans²,

Wagenmakers³

2022

Psychological Methods

Self Cite

View full text Add to dashboard Cite

The Bayesian statistical framework requires the specification of prior distributions, which reflect predata knowledge about the relative plausibility of different parameter values. As prior distributions influence the results of Bayesian analyses, it is important to specify them with care. Prior elicitation has frequently been proposed as a principled method for deriving prior distributions based on expert knowledge. Although prior elicitation provides a theoretically satisfactory method of specifying prior distributions, there are several implicit decisions that researchers need to make at different stages of the elicitation process, each of them constituting important researcher degrees of freedom. Here, we discuss some of these decisions and group them into 3 categories: decisions about (a) the setup of the prior elicitation; (b) the core elicitation process; and (c) combination of elicited prior distributions from different experts. Importantly, different decision paths could result in greatly varying priors elicited from the same experts. Hence, researchers who wish to perform prior elicitation are advised to carefully consider each of the practical decisions before, during, and after the elicitation process. By explicitly outlining the consequences of these practical decisions, we hope to raise awareness for methodological flexibility in prior elicitation and provide researchers with a more structured approach to navigate the decision paths in prior elicitation. Making the decisions explicit also provides the foundation for further research that can identify evidence-based best practices that may eventually reduce the methodologically flexibility in prior elicitation.

show abstract

“…To test for specific fixed effects in linear mixed models, we obtained p-values using the Kenward-Roger method, and BFs through Bayesian model averaging by estimating the change from prior to posterior inclusion odds (inclusion BF). In other words, this model-averaged BF indicates how much more likely the data are under model variants that include a given fixed effect, compared to model variants that exclude the fixed effect (Hinne et al ., 2020).…”

Section: Methodsmentioning

confidence: 99%

Locus coeruleus integrity and the effect of atomoxetine on response inhibition in Parkinson’s disease

O’Callaghan

Hezemans

et al. 2020

Preprint

View full text Add to dashboard Cite

Cognitive decline is a common feature of Parkinson's disease, and many of these cognitive deficits fail to respond to dopaminergic therapy. Therefore, targeting other neuromodulatory systems represents an important therapeutic strategy. Among these, the locus coeruleus-noradrenaline system has been extensively implicated in response inhibition deficits. Restoring noradrenaline levels using the noradrenergic reuptake inhibitor atomoxetine can improve response inhibition in some patients with Parkinson's disease, but there is considerable heterogeneity in treatment response. Accurately predicting the patients who would benefit from therapies targeting this neurotransmitter system remains a critical goal, in order to design the necessary clinical trials with stratified patient selection to establish the therapeutic potential of atomoxetine. Here, we test the hypothesis that integrity of the noradrenergic locus coeruleus explains the variation in improvement of response inhibition following atomoxetine. In a double-blind placebo-controlled randomised crossover design, 19 people with Parkinson's disease completed an acute psychopharmacological challenge with 40 mg of oral atomoxetine or placebo. A stop-signal task was used to measure response inhibition, with stop-signal reaction times obtained through hierarchical Bayesian estimation of an ex-Gaussian race model. Twenty-six control subjects completed the same task without undergoing the drug manipulation. In a separate session, patients and controls underwent ultra-high field 7T imaging of the locus coeruleus using a neuromelanin-sensitive magnetisation transfer sequence. The principal result was that atomoxetine improved stop-signal reaction times in those patients with lower locus coeruleus integrity. This was in the context of a general impairment in response inhibition, as patients on placebo had longer stop-signal reaction times compared to controls. We also found that the caudal portion of the locus coeruleus showed the largest neuromelanin signal decrease in the patients compared to controls. Our results highlight a link between the integrity of the noradrenergic locus coeruleus and response inhibition in Parkinson's disease patients. Furthermore, they demonstrate the importance of baseline noradrenergic state in determining the response to atomoxetine. We suggest that locus coeruleus neuromelanin imaging offers a marker of noradrenergic capacity that could be used to stratify patients in trials of noradrenergic therapy and to ultimately inform personalised treatment approaches.

show abstract

A conceptual introduction to Bayesian Model Averaging

Cited by 25 publications

References 36 publications

Selection in working memory is resource-demanding: Concurrent task effects on the retro-cue effect

Selection in working memory is resource-demanding: Concurrent task effects on the retro-cue effect

Practical challenges and methodological flexibility in prior elicitation.

Locus coeruleus integrity and the effect of atomoxetine on response inhibition in Parkinson’s disease

Contact Info

Product

Resources

About