Calibrate scores under measurement error

Usage

rccme_calib_me(
  w_mat,
  rel_vec = NULL,
  w_se_vec = NULL,
  w_se_mat = NULL,
  z_mat = matrix(0, nrow = nrow(w_mat), ncol = 0),
  rescale = TRUE,
  standard = FALSE
)

Arguments

w_mat: (matrix) a matrix of trait estimates for n respondents on p latent variables
rel_vec: (vector) a vector of marginal reliability for p latent variables
w_se_vec: (vector) a vector of standard errors of measurement for p latent variables
w_se_mat: (matrix) a matrix of trait estimates standard-errors for n respondents on p latent variables
z_mat: (matrix) a matrix of error-free covariates for n respondents and q covariates
rescale: (logical) Should the trait estimates and their standard errors be re-scaled? Default is TRUE. The variables are rescaled under the assumption that the trait estimates were not conditioned on any background variables and that the marginal variance of the latent trait is 1. This is true for standardised latent variables in CFA or the default prior variance assumption of 1 in IRT score estimates. This rescaling ensures the trait coefficients are correct for the standardised traits.
standard: (logical) Only relevant when passing reliability. If TRUE, attempt to return the standardised version of the calibrated scores If FALSE (default), do not attempt standardisation.

Value

Calibrated trait estimates.