make-up mini-quiz ARTICLE

Question 1

What is the primary goal of joint modeling?

Answer 1

To model the covariation between parameters of neural and behavioral submodels, providing an integrated understanding of cognitive processes.

Question 2

Name the three types of joint models discussed in the document.

Answer 2

Integrative, Directed, and Covariance joint models.

These approaches provide flexibility in modeling depending on the nature of the data and the research goals.

Question 3

Explain what Integrative Joint Models are.

Answer 3

explain both neural and behavioral data simultaneously
using single set of parameters
require strong theoretical commitments
powerful but challenging to develop

Question 4

Explain what Directed Joint Models are.

Answer 4

link neural to behavioral parameters (or vice versa)
uses deterministic mapping function
The parameters directly modulate each other
offers flexibility in handling mismatched data scales

Question 5

Explain what Covariance Joint Models are.

Answer 5

neural and behavioral parameters are related through probabilistic methods
uses shared variance-covariance structure
less constrained by deterministic links
handle data variability better
computationally more complex and require large datasets.

Question 6

What is Marr’s hierarchy, and which levels do mathematical psychologists and cognitive neuroscientists typically focus on?

Answer 6

computational, algorithmic, and implementational
Mathematical psychologists focus on computational and algorithmic levels, while cognitive neuroscientists focus on the implementational level.

Question 7

What is Bayesian modeling, and why is it useful in joint modeling?

Answer 7

updates probability of hypothesis as more evidence comes in
good for integrating uncertainty
and making inferences about complex relationships

Question 8

What is the purpose of the Just Another Gibbs Sampler (JAGS)?

Answer 8

JAGS is software used for Bayesian inference by sampling from posterior distributions in complex models.

Question 9

Define the ‘Integrative’ joint modeling approach.

Answer 9

In the Integrative approach, a single set of parameters explains both neural and behavioral data streams simultaneously.

Question 10

What is the key difference between Directed and Covariance joint models?

Answer 10

Directed: neural parameters to modulate behavioral directly
Covariance: relate parameters through shared statistical distribution

Question 11

What is the logistic function used for in the behavioral sub-model?

Answer 11

It transforms a latent familiarity parameter into a probability of an “old” response in recognition memory tasks.

Question 12

Why are multivariate normal distributions often used in Covariance models?

Answer 12

They conveniently describe relationships between parameters, including central tendencies and variances/covariances.

Question 13

What is the purpose of the hyperparameters 𝜙 and Σ in Covariance models?

Answer 13

They represent the mean vector and variance-covariance matrix governing the relationship between neural and behavioral parameters.

Question 14

List two advantages of Covariance joint models over Directed models.

Answer 14

1. Better handling of outliers and missing data
2. Flexibility in specifying probabilistic relationships between parameters.

Question 15

What is the main limitation of the Integrative joint modeling approach?

Answer 15

It requires strong commitments about cognitive processes and their neural correlates, making it theoretically and computationally challenging.

Question 16

How does the Directed joint model handle mismatched data sizes between neural and behavioral streams?

Answer 16

It uses linking functions that map neural parameters to behavioral parameters, allowing flexibility despite differences in data scales.

Question 17

What does a Hidden Markov Model (HMM) help achieve in joint modeling?

Answer 17

It helps model latent states and their transitions, particularly in scenarios with temporal neural and behavioral data.

Question 18

What is the role of priors in Bayesian joint modeling?

Answer 18

Priors encode initial beliefs about model parameters and contribute to forming posterior distributions when combined with data.

Question 19

Explain why sampling methods, like Gibbs sampling, are necessary for joint models.

Answer 19

Because joint models often involve complex and high-dimensional posterior distributions that cannot be solved analytically.

Question 20

What is the purpose of simulating data when building a joint model?

Answer 20

To test and validate the model’s ability to accurately estimate parameters and fit the data.

Question 21

What are ‘plates’ in graphical representations of joint models?

Answer 21

Plates indicate replication of nodes across dimensions, such as trials or time points, in the model.

Question 22

How do Covariance models mitigate the impact of outliers in neural data?

Answer 22

By modeling parameters as latent variables governed by overarching distributions, reducing direct effects of outliers on parameter estimates.

Question 23

What is the significance of using real-world data in the tutorial’s final example?

Answer 23

To demonstrate how joint modeling techniques can be applied to actual experimental data for meaningful scientific inference.

Question 24

What is the main advantage of using Bayesian change point analysis in joint models?

Answer 24

It helps identify points in time where the statistical properties of data (e.g., means or variances) change, which is useful for modeling time-varying processes.

Question 25

What does the term ‘parent-to-child’ ancestry refer to in joint model diagrams?

Answer 25

It indicates dependency between variables, where the child node’s value depends on the parent node’s value.

Question 26

Why are outliers less impactful in Covariance models compared to Directed models?

Answer 26

Covariance models treat parameters as latent variables and rely on overarching distributions, which absorb the variability caused by outliers.

Question 27

What is the purpose of the inverse Wishart distribution in the model?

Answer 27

It serves as a prior for the variance-covariance matrix in Bayesian modeling, ensuring conjugacy for efficient sampling.

Question 28

What is the relationship between the hyperparameters 𝜙 and Σ in the Covariance approach?

Answer 28

𝜙 represents the mean vector of the parameters, while Σ defines their variance-covariance structure.

Question 29

What is the practical limitation of the Covariance approach concerning data requirements?

Answer 29

It requires a large amount of data to estimate parameters accurately due to its complexity and multiple levels of parameters.

Question 30

What role do regression parameters (𝛽𝑘) play in Directed models?

Answer 30

They map neural parameters to behavioral parameters, enabling prediction and modeling of relationships between the two.

Question 31

What is the significance of the drift rate parameter in decision models, as used in the document’s example?

Answer 31

It represents the speed at which evidence accumulates toward a decision threshold, linking neural activity to behavior.

Question 32

What are the advantages of joint modelling over a two-stage correlation approach?

Answer 32

joint models are better equipped to (1) handle mismatching (i.e., when the size of the neural data is different from the size of the behavioral data) and missing data, (2) perform inference on the magnitude of brain-behavior relationships (i.e., they are not subject to Type I errors as in the two-stage approach), (3) compare different brain-behavior relationships across models, and (4) make predictions about either neural or behavioral data.