mlm.spike.slab.prior {BoomSpikeSlab}R Documentation

Create a spike and slab prior for use with mlm.spike.

Description

Creates a spike and slab prior for use with mlm.spike.

Usage

MultinomialLogitSpikeSlabPrior(
    response,
    subject.x,
    expected.subject.model.size = 1,
    choice.x = NULL,
    expected.choice.model.size = 1,
    max.flips = -1,
    nchoices = length(levels(response)),
    subject.dim = ifelse(is.null(subject.x), 0, ncol(subject.x)),
    choice.dim = ifelse(is.null(choice.x), 0, ncol(choice.x)))

Arguments

response

The response variable in the multinomial logistic regression. The response variable is optional if nchoices is supplied. If 'response' is provided then the prior means for the subject level intercpets will be chosen to match the empirical values of the response.

subject.x

The design matrix for subject-level predictors. This can be NULL or of length 0 if no subject-level predictors are present.

expected.subject.model.size

The expected number of non-zero coefficients – per choice level – in the subject specific portion of the model. All coefficients can be forced into the model by setting this to a negative number, or by setting it to be larger than the dimension of the subject-level predictors.

choice.x

The design matrix for choice-level predictors. Each row of this matrix represents the characteristics of a choice in a choice occasion, so it takes 'nchoices' rows to encode one observation. This can be NULL or of length 0 if no choice-level predictors are present.

expected.choice.model.size

The expected number of non-zero coefficients in the choice-specific portion of the model. All choice coefficients can be forced into the model by setting this to a negative number, or by setting it to be larger than the dimension of the choice-level predictors (for a single response level).

max.flips

The maximum number of variable inclusion indicators the sampler will attempt to sample each iteration. If max.flips <= 0 then all indicators will be sampled.

nchoices

Tne number of potential response levels.

subject.dim

The number of potential predictors in the subject-specific portion of the model.

choice.dim

The number of potential predictors in the choice-specific portion of the model.

Value

An object of class IndependentSpikeSlabPrior, with elements arranged as expected by mlm.spike.

Author(s)

Steven L. Scott

References

Tuchler (2008), "Bayesian Variable Selection for Logistic Models Using Auxiliary Mixture Sampling", Journal of Computational and Graphical Statistics, 17 76 – 94.


[Package BoomSpikeSlab version 1.2.4 Index]