Determining informative priors for cognitive models

Lee, Michael; Vanpaemel, Wolf

doi:10.3758/s13423-017-1238-3

Cited by 119 publications

(129 citation statements)

References 71 publications

Supporting

Mentioning

125

Contrasting

Order By: Relevance

“…In practice, Lee and Vanpaemel () argue, it is reasonable to develop priors based not only on guiding theory, but on previous values from other relevant data, or from considering the reasonableness of the predictions about data that a choice of priors implies. Accordingly, the priors in Figure allow for sensitivity, bias, and the standard deviation to take a range of values consistent with theoretical expectations, and previous modeling results .…”

Section: Bayesian Modeling Analysesmentioning

confidence: 99%

Bayesian methods applied to the generalized matching law

Villarreal

Velázquez

Baroja

et al. 2019

J Exper Analysis Behavior

Self Cite

View full text Add to dashboard Cite

We demonstrate the usefulness of Bayesian methods in developing, evaluating, and using psychological models in the experimental analysis of behavior. We do this through a case study, involving new experimental data that measure the response count and time allocation behavior in pigeons under concurrent random‐ratio random‐interval schedules of reinforcement. To analyze these data, we implement a series of behavioral models, based on the generalized matching law, as graphical models, and use computational methods to perform fully Bayesian inference. We demonstrate how Bayesian methods, implemented in this way, make inferences about parameters representing psychological variables, how they test the descriptive adequacy of models as accounts of behavior, and how they compare multiple competing models. We also demonstrate how the Bayesian graphical modeling approach allows for more complicated modeling structures, including hierarchical, common cause, and latent mixture structures, to formalize more complicated behavioral models. As part of the case study, we demonstrate how the statistical properties of Bayesian methods allow them to provide more direct and intuitive tests of theories and hypotheses, and how they support the creative and exploratory development of new theories and models.

show abstract

Section: Bayesian Modeling Analysesmentioning

confidence: 99%

Bayesian methods applied to the generalized matching law

Villarreal

Velázquez

Baroja

et al. 2019

J Exper Analysis Behavior

Self Cite

View full text Add to dashboard Cite

show abstract

“…In response, others have argued that priors placed on model parameters are a vital component of the theory, and therefore model selection should be sensitive to the priors . Theoretical development could reflect different theoretical assumptions of the priors on θ and model selection could be performed over these different instantiations.…”

Section: Bayesian Model Selectionmentioning

confidence: 99%

“…64 For others, an informative prior is viewed as an integral part of the model, forcing the theorist to formalize assumptions about model parameters. 65,66 In the case of cognitive psychology, we often have too little prior information to set informative priors in a way that would be universally accepted, so we often use noninformative priors. There is a large body of research devoted to developing noninformative priors which can be used as (at least arguably) a reasonable default.…”

Section: Stimulus Onsetmentioning

confidence: 99%

Bayesian statistical approaches to evaluating cognitive models

Annis

Palmeri

2017

WIRES Cognitive Science

View full text Add to dashboard Cite

Cognitive models aim to explain complex human behavior in terms of hypothesized mechanisms of the mind. These mechanisms can be formalized in terms of mathematical structures containing parameters that are theoretically meaningful. For example, in the case of perceptual decision making, model parameters might correspond to theoretical constructs like response bias, evidence quality, response caution, and the like. Formal cognitive models go beyond verbal models in that cognitive mechanisms are instantiated in terms of mathematics and they go beyond statistical models in that cognitive model parameters are psychologically interpretable. We explore three key elements used to formally evaluate cognitive models: parameter estimation, model prediction, and model selection. We compare and contrast traditional approaches with Bayesian statistical approaches to performing each of these three elements. Traditional approaches rely on an array of seemingly ad hoc techniques, whereas Bayesian statistical approaches rely on a single, principled, internally consistent system. We illustrate the Bayesian statistical approach to evaluating cognitive models using a running example of the Linear Ballistic Accumulator model of decision making (Brown SD, Heathcote A. The simplest complete model of choice response time: linear ballistic accumulation. Cogn Psychol 2008, 57:153-178). WIREs Cogn Sci 2018, 9:e1458. doi: 10.1002/wcs.1458 This article is categorized under: Neuroscience > Computation Psychology > Reasoning and Decision Making Psychology > Theory and Methods.

show abstract

“…The behavioral model we choose can be anything, such as a signal detection theory model for data from a perceptual discrimination experiment [31][32][33][34][35][36][37][38][39][40], the bind cue decide model of episodic memory [41], or the drift diffusion model for choice response times (see [42] and Chap. 15 of this text).…”

Section: Joint Modelingmentioning

confidence: 99%

“…The chapter elaborates on the work of Turner et al [23], where the authors conjoin "submodels" of singular (i.e., neural or behavioral, but not both) measures by way of a hierarchical Bayesian framework. Bayesian modeling has become popular in many neural [24][25][26][27][28][29] and behavioral [30][31][32][33][34][35][36][37][38][39] modeling applications for a number of theoretical and practical reasons (see Chap. 9 of this text).…”

Section: Introductionmentioning

confidence: 99%

Constraining Cognitive Abstractions Through Bayesian Modeling

Turner

2015

An Introduction to Model-Based Cognitive Neuroscience

View full text Add to dashboard Cite

There are many ways to combine neural and behavioral measures to study cognition. Some ways are theoretical, and other ways are statistical. The predominant statistical approach treats both sources of data as independent and the relationship between the two measures is inferred by way of a (post hoc) regression analysis. In this chapter, we review an alternative approach that allows for flexible modeling of both measures simultaneously. We then explore and elaborate on several of the most important benefits of this modeling approach, and close with a model comparison of the Linear Ballistic Accumulator model and a drift diffusion model on neural and behavioral data. IntroductionAs this book will demonstrate, there are many ways in which the neurosciences can inspire mathematical models of cognition, and vice versa. Perhaps the most theoretically-oriented way is to develop models on the basis of what is observed at the neurophysiological level. This approach could be called "bottom-up" because at its very core, the models are designed to embody neurological principles [1][2][3][4][5][6][7][8][9].If a model is not originally designed as a bottom-up model, it is possible to adjust the assumptions of the model so that it resembles a bottom-up structure. An example of this is the latest instantiation of the ACT-R model (see [10] and Chap. 18 of this text). Anderson and colleagues [11] developed a model of the process of equation solving that also predicted patterns of brain activity, as measured by the blood oxygen level dependent (BOLD) response. Their model assumed that observers maintain a set of modules that become active when required by subtask demands, and become inactive when no longer in use. In the model, BOLD responses are produced in the brain areas corresponding to active modules by convolving a module activation function (i.e., a function carrying binary active or inactive information) with a hemodynamic response function. The correspondence between modules and B. M. Turner ( )

show abstract

Determining informative priors for cognitive models

Cited by 119 publications

References 71 publications

Bayesian methods applied to the generalized matching law

Bayesian methods applied to the generalized matching law

Bayesian statistical approaches to evaluating cognitive models

Constraining Cognitive Abstractions Through Bayesian Modeling

Contact Info

Product

Resources

About