A list of parameters for controlling the prior distributions. These include:
type: Vector of four strings indicating the type of prior distributions to use. In order, these are: 1) priors for all column-specific intercepts, row effects, and cutoff points for ordinal data; 2) priors for the latent variable coefficients and correlation parameters. This is ignored for this function; 3) priors for all column-specific coefficients relating to X (ignored if X = NULL). When traits are included in the model, this is also the prior for the trait regression coefficients (please see about.traits for more information); 4) priors for any dispersion parameters and variance (standard deviation, to be precise) parameters in the model.
For elements 1-3, the prior distributions currently available include: I) ``normal", which is a normal prior with the variance controlled by elements 1-3 in hypparams; II) ``cauchy", which is a Cauchy prior with variance controlled by elements 1-3 in hypparams. Gelman, et al. (2008) considers using Cauchy priors with variance \(2.5^2\) as weakly informative priors for coefficients in logistic and potentially other generalized linear models; III) ``uniform", which is a symmetric uniform prior with minimum and maximum values controlled by element 1-3 in hypparams.
For element 4, the prior distributions currently available include: I) ``uniform", which is uniform prior with minimum zero and maximum controlled by element 4 in hypparmas; II) ``halfnormal", which is half-normal prior with variance controlled by hypparams; III) ``halfcauchy", which is a half-Cauchy prior with variance controlled by element 4 in hypparams.
Defaults to the vector c("normal","normal","normal","uniform").
hypparams Vector of four hyperparameters used in the set up of prior distributions. In order, these are: 1) affects the prior distribution for all column-specific intercepts, row effects, and cutoff points for ordinal data; 2) affects the prior distribution for all latent variable coefficients and correlation parameters. This is ignored for this function; 3) affects the prior distribution for column-specific coefficients relating to X (ignored if X = NULL). When traits are included in the model, it also affects the prior distribution for the trait regression coefficients; 4) affects the prior distribution for any dispersion parameters, as well as the prior distributions for the standard deviation of the random effects normal distribution if row.eff = "random", the standard deviation of the column-specific random intercepts for these columns if more than two of the columns are ordinal, and the standard deviation of the random effects normal distribution for trait regression coefficients when traits are included in the model.
Defaults to the vector c(10, 10, 10, 30). The use of normal distributions with mean zero and variance 10 as priors is seen as one type of (very) weakly informative prior, according to Prior choice recommendations.
ssvs.index: Indices to be used for stochastic search variable selection (SSVS, George and McCulloch, 1993). Either a single element or a vector with length equal to the number of columns in the implied model matrix X. Each element can take values of -1 (no SSVS is performed on this covariate), 0 (SSVS is performed on individual coefficients for this covariate), or any integer greater than 0 (SSVS is performed on collectively all coefficients on this covariate/s.)
Please see about.ssvs for more information regarding the implementation of SSVS. Defaults to -1, in which case SSVS is not performed on X variables.
ssvs.g: Multiplicative, shrinkage factor for SSVS, which controls the strength of the "spike" in the SSVS mixture prior. In summary, if the coefficient is included in the model, the "slab" prior is a normal distribution with mean zero and variance given by element 3 in hypparams, while if the coefficient is not included in the model, the "spike" prior is normal distribution with mean zero and variance given by element 3 in hypparams multiplied by ssvs.g. Please see about.ssvs for more information regarding the implementation of SSVS. Defaults to 1e-6.
ssvs.traitsindex: Used in conjunction with traits and which.traits, this is a list of indices to be used
for performing SSVS on the trait coefficients. Should be a list with the same length as which.traits, and with each element a vector of indices with the same length as the corresponding element in which.traits. Each index either can take values of -1 (no SSVS on this trait coefficient) or 0 (no SSVS on this trait coefficient).
Please see about.ssvs for more information regarding the implementation of SSVS. Defaults to -1, in which case SSVS is not performed on any of the trait coefficients, if they are included in the model.