group: Group or cluster covariates

In andrewzm/INLA: Functions which allow to perform full Bayesian analysis of latent Gaussian models using Integrated Nested Laplace Approximaxion

Description

inla.group group or cluster covariates so to reduce the number of unique values

Usage

inla.group(x, n = 25, method = c("cut", "quantile"), idx.only = FALSE)

Arguments

x	The vector of covariates to group.
n	Number of classes or bins to group into.
method	Group either using bins with equal length intervals (method = "cut"), or equal distance in the ‘probability’ scale using the quantiles (method = "quantile").
idx.only	Option to return the index only and not the method.

Value

inla.group return the new grouped covariates where the classes are set to the median of all the covariates belonging to that group.

Author(s)

Havard Rue [e-mail]

See Also

f

Examples

## this gives groups 3 and 8
x = 1:10
x.group = inla.group(x, n = 2)

## this is the intended use, to reduce the number of unique values in
## the of first argument of f()
n = 100
x = rnorm(n)
y = x + rnorm(n)
result = inla(y ~ f(inla.group(x, n = 20), model = "iid"), data=data.frame(y=y,x=x))

andrewzm/INLA documentation built on May 10, 2019, 11:12 a.m.