nominalmod: Estimation of a nominal response model

In micbtz/regIRT: Regularized estimation of IRT models

Description

Estimates the parameters of the nominal response model with optional penalty on the slope parameters.

Usage

nominalmod(data, D, parini, parW = NULL, lambda = 0, pen = NULL, adaptive = NULL, 
  items.select = 1:ncol(data), nq = NULL)

Arguments

data	dataset.
D	number of dimensions.
parini	initial values for the parameters.
parW	vector of parameters used for computing weights for the adaptive penalization.
lambda	vector of tuning parameters.
pen	type of penalization: "lasso" or "ridge".
adaptive	logical; if TRUE adaptive lasso is performed.
items.select	vector of integer values indicating the items with penalty.
nq	number of quadrature points per dimension. By default the number of quadrature points depends on the number of dimensions: '1'=61, '2'=31, '3'=15, '4'=9, '5'=7, '>5'=3.

Details

If lambda is zero, no penalitazion is applied. If lambda contains more elements, the model is estimated for each value.

Value

A list with components:

data	dataset.
D	number of dimensions.
parini	initial values for the parameters.
parW	vector of parameters used for computing weights for the adaptive penalization.
lambda	vector of tuning parameters.
pen	type of penalization: "lasso" or "ridge".
adaptive	logical; if TRUE adaptive lasso is performed.
items.select	vector of integer values indicating the items with penalty.
nq	number of quadrature points per dimension.
par	matrix of parameter estimates. Columns correspond to different values of lambda.
lik	vector containing the penalized log-likelihood computed for each lambda.
convergence	An integer code. 0 indicates successful completion.

See Also

nominalCV, regIRT, regPath

micbtz/regIRT documentation built on July 6, 2019, 2:37 p.m.