Intuitive Construction of Joint Priors for Variance Parameters
Internal variance parameter order
List of available plotting functions
Extract the posterior of a random effect
Extract the posterior parameter estimate
Find suitable PC prior parameters
Compile stan-model
Create a "skeleton" for custom Stan code
Evaluate the joint variance prior
Evaluate PC prior for variance proportion
expit
Run inference
Run inference
logit
Making a prior object
Returning a simple example prior object
Graphical prior construction
List available priors, latent models and likelihoods
Define latent component
Plotting prior for a single parameter (weight or variance (not standar...
Plotting posterior distributions
Plotting posterior distributions
Plotting posterior variances, standard deviations or precisions
Plotting prior distributions
Plotting several posterior distributions
Plotting the prior tree structure graph
Plotting
Scaling precision matrix
Short summary
Compute the typical variance
Tool for easy prior construction and visualization. It helps to formulates joint prior distributions for variance parameters in latent Gaussian models. The resulting prior is robust and can be created in an intuitive way. A graphical user interface (GUI) can be used to choose the joint prior, where the user can click through the model and select priors. An extensive guide is available in the GUI. The package allows for direct inference with the specified model and prior. Using a hierarchical variance decomposition, we formulate a joint variance prior that takes the whole model structure into account. In this way, existing knowledge can intuitively be incorporated at the level it applies to. Alternatively, one can use independent variance priors for each model components in the latent Gaussian model. Details can be found in the accompanying scientific paper: Hem, Fuglstad, Riebler (2024, Journal of Statistical Software, <doi:10.18637/jss.v110.i03>).