Dataset Information

Forget-me-some: General versus special purpose models in a hierarchical probabilistic task.

ABSTRACT: Humans build models of their environments and act according to what they have learnt. In simple experimental environments, such model-based behaviour is often well accounted for as if subjects are ideal Bayesian observers. However, more complex probabilistic tasks require more sophisticated forms of inference that are sufficiently computationally and statistically taxing as to demand approximation. Here, we study properties of two approximation schemes in the context of a serial reaction time task in which stimuli were generated from a hierarchical Markov chain. One, pre-existing, scheme was a generically powerful variational method for hierarchical inference which has recently become popular as an account of psychological and neural data across a wide swathe of probabilistic tasks. A second, novel, scheme was more specifically tailored to the task at hand. We show that the latter model fit significantly better than the former. This suggests that our subjects were sensitive to many of the particular constraints of a complex behavioural task. Further, the tailored model provided a different perspective on the effects of cholinergic manipulations in the task. Neither model fit the behaviour on more complex contingencies that competently. These results illustrate the benefits and challenges that come with the general and special purpose modelling approaches and raise important questions of how they can advance our current understanding of learning mechanisms in the brain.

SUBMITTER: Broker F

PROVIDER: S-EPMC6197684 | biostudies-literature | 2018

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Forget-me-some: General versus special purpose models in a hierarchical probabilistic task.

Bröker Franziska F Marshall Louise L Bestmann Sven S Dayan Peter P

PloS one 20181022 10

Humans build models of their environments and act according to what they have learnt. In simple experimental environments, such model-based behaviour is often well accounted for as if subjects are ideal Bayesian observers. However, more complex probabilistic tasks require more sophisticated forms of inference that are sufficiently computationally and statistically taxing as to demand approximation. Here, we study properties of two approximation schemes in the context of a serial reaction time ta ...[more]

PMID: 30346977

Similar Datasets

Project description:[Correction Notice: An Erratum for this article was reported in Vol 129(7) of Journal of Abnormal Psychology (see record 2020-72912-001). In the article (http://dx.doi.org/10.1037/abn0000601), an acknowledgment is missing from the author note. The missing acknowledgement is included in the erratum.] Psychopathology can be viewed as a hierarchy of correlated dimensions. Many studies have supported this conceptualization, but they have used alternative statistical models with differing interpretations. In bifactor models, every symptom loads on both the general factor and 1 specific factor (e.g., internalizing), which partitions the total explained variance in each symptom between these orthogonal factors. In second-order models, symptoms load on one of several correlated lower-order factors. These lower-order factors load on a second-order general factor, which is defined by the variance shared by the lower-order factors. Thus, the factors in second-order models are not orthogonal. Choosing between these valid statistical models depends on the hypothesis being tested. Because bifactor models define orthogonal phenotypes with distinct sources of variance, they are optimal for studies of shared and unique associations of the dimensions of psychopathology with external variables putatively relevant to etiology and mechanisms. Concerns have been raised, however, about the reliability of the orthogonal specific factors in bifactor models. We evaluated this concern using parent symptom ratings of 9-10 year olds in the ABCD Study. Psychometric indices indicated that all factors in both bifactor and second-order models exhibited at least adequate construct reliability and estimated replicability. The factors defined in bifactor and second-order models were highly to moderately correlated across models, but have different interpretations. All factors in both models demonstrated significant associations with external criterion variables of theoretical and clinical importance, but the interpretation of such associations in second-order models was ambiguous due to shared variance among factors. (PsycInfo Database Record (c) 2020 APA, all rights reserved).

Dataset Information

Forget-me-some: General versus special purpose models in a hierarchical probabilistic task.

Publications

Forget-me-some: General versus special purpose models in a hierarchical probabilistic task.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets