loss_function: loss-function with penalization

In submission-user/cmlDIFlasso: DIF-detection with Lasso-methods in a cML-Framework

Description

calculates logclik with or without penalization for DIF-model given data, parameter values for items, DIF-parameters per variable per item and value of lambda

Usage

loss_function(
  data,
  covariates,
  beta_intercept = rep(0, ncol(data)),
  delta_matrix = matrix(0, ncol(covariates), ncol(data)),
  DIFmodel = "RM",
  restr_pen = FALSE,
  lambda = 0.5,
  lambda_restr = 100
)

Arguments

data	Answers of all persons (0/1, n x I)
covariates	Matrix of covariate values of all persons (n x ncovs)
beta_intercept	vecotr of parameter values for intercepts per item
delta_matrix	Matrix (ncovs x nitems) of DIF-parameters per variable and item
DIFmodel	can be "items", "variables", "all.interactions", "none". And FALSE for Rasch model estimation
restr_pen	For additional penalization to ensure zero sum of intercepts
lambda	penalization parameter
lambda_restr	penalization parameter for restriction

Value

the value of the (penalized) log conditional Likelihood

submission-user/cmlDIFlasso documentation built on Dec. 23, 2021, 6:40 a.m.