FitGLLVM: Fits a GLLVM to data

In oharar/LatentINLA: Package to Write GLLVMs in INLA

Fits a GLLVM to data

Description

Fits a GLLVM to data

Usage

FitGLLVM(
  Y,
  X = NULL,
  W = NULL,
  nLVs = 1,
  Family = "gaussian",
  RowEff = "fixed",
  ColEff = "fixed",
  RowEffPriorsd = 100,
  ColEffPriorsd = 100,
  ColScorePriorsd = 10,
  PriorLV = NULL,
  INLAobj = FALSE,
  ...
)

Arguments

Y	A data frame or matrix with the response (assuming counts at the moment)
X	A data frame or matrix of covariates (strictly row-level covariates). Can be NULL
W	A data frame or matrix of column-level covariates. Can be NULL
nLVs	The number of latent variables required
Family	A string indicating the likelihood family. If length 1, it gets repeated with one for each column of the data. For supported distributions see names(inla.models()$likelihood).
RowEff	String indicating what sort of row effect is required. Either none, fixed or random. Defaults to fixed.
ColEff	String indicating what sort of column effect is required. Either none, fixed or random. Defaults to fixed.
RowEffPriorsd	Prior standard deviation for latent variable, defaults to 100
ColEffPriorsd	Prior standard deviation for latent variable, defaults to 100
ColScorePriorsd	Prior standard deviation for column scores (the betas for INLA insiders), defaults to 10
PriorLV	Hyperprior for the precision of the latent variable, as a list that INLA will understand (sorry). Defaults to NULL, where the default INLA prior will be used
INLAobj	Should the full INLA object be included in the output object? Defaults to FALSE
...	More arguments to be passed to inla()

Details

This fits a GLLVM....

Priors for fixed effects can be passed straight to INLA with code like control.fixed = list(mean = ..., prec = ...)

The column effects have to be fixed effects: if you want them to be random, you have to have fixed rows. You can just transpose the matrix to get this model.

Value

A list with fixed, rowterm, colterm, colscores, and roweffs, formula, Y, X, family. the posterior summaries for the fixed effects, row main effects, the column main effect, the column scores and the row effects respectively. rowterm and colterm may be NULL if they were not set to be random.

Examples

FitGLLVM(matrix(1:10, ncol=5), nLVs=1, Family="poisson")

oharar/LatentINLA documentation built on Sept. 13, 2023, 5:18 p.m.