PartInvMulti_we: Evaluate selection accuracy based on the MCAA Framework

In claycantrell/PartInvShinyUI: Multidimensional Classification Accuracy Analysis Framework

Evaluate selection accuracy based on the MCAA Framework

Description

PartInv, PartInvMulti_we evaluate partial measurement invariance using an extension of Millsap & Kwok's (2004) approach

Usage

PartInvMulti_we(
  propsel,
  cut_z = NULL,
  weights_item = NULL,
  weights_latent = NULL,
  kappa_r = NULL,
  kappa_f = kappa_r,
  alpha_r,
  alpha_f = alpha_r,
  phi_r = NULL,
  phi_f = phi_r,
  psi_r,
  psi_f = psi_r,
  lambda_r,
  lambda_f = lambda_r,
  tau_r = NULL,
  tau_f = tau_r,
  nu_r,
  nu_f = nu_r,
  Theta_r,
  Theta_f = Theta_r,
  pmix_ref = 0.5,
  plot_contour = TRUE,
  labels = c("Reference", "Focal"),
  ...
)

PartInv(
  propsel,
  cut_z = NULL,
  weights_item = NULL,
  weights_latent = NULL,
  kappa_r = NULL,
  kappa_f = kappa_r,
  alpha_r,
  alpha_f = alpha_r,
  phi_r = NULL,
  phi_f = phi_r,
  psi_r,
  psi_f = psi_r,
  lambda_r,
  lambda_f = lambda_r,
  tau_r = NULL,
  tau_f = tau_r,
  nu_r,
  nu_f = nu_r,
  Theta_r,
  Theta_f = Theta_r,
  pmix_ref = 0.5,
  plot_contour = TRUE,
  labels = c("Reference", "Focal"),
  ...
)

Arguments

propsel	proportion of selection. If missing, computed using cut_z.
cut_z	pre-specified cutoff score on the observed composite. This argument is ignored when propsel has input.
weights_item	a vector of item weights
weights_latent	a vector of latent factor weights
alpha_r	a vector of latent factor means for the reference group.
alpha_f	(optional) a vector of latent factor means for the focal group; if no input, set equal to alpha_r.
phi_r, phi_f, tau_r, tau_f, kappa_r, kappa_f	deprecated; included only for backward compatibility.
psi_r	a matrix of latent factor variance-covariances for the reference group.
psi_f	(optional) a matrix of latent factor variance-covariances for the focal group; if no input, set equal to psi_r.
lambda_r	a matrix of factor loadings for the reference group.
lambda_f	(optional) a matrix of factor loadings for the focal group; if no input, set equal to lambda_r.
nu_r	a matrix of measurement intercepts for the reference group.
nu_f	(optional) a matrix of measurement intercepts for the focal group; if no input, set equal to nu_r.
Theta_r	a matrix of the unique factor variances and covariances for the reference group.
Theta_f	(optional) a matrix of the unique factor variances and covariances for the focal group; if no input, set equal to Theta_r.
pmix_ref	Proportion of the reference group; default to 0.5 (i.e., two populations have equal size).
plot_contour	logical; whether the contour of the two populations should be plotted; default to TRUE.
labels	a character vector with two elements to label the reference and the focal group on the graph.
...	other arguments passed to the contour function.

Value

The output will be a list of four elements and a plot if plot_contour == TRUE:

1. propsel: echo the same argument as input.

2. cutpt_xi: cut point on the latent scale (xi).

3. cutpt_z: cut point on the observed scale (Z).

4. summary: A 8 x 3 table, with columns representing the reference, the focal, and the expected results if the latent distribution of focal group matches the reference group. The rows represent probabilities of true positive (A), false positive (B), true negative (C), false negative (D); proportion selected, success ratio, sensitivity, and specificity.

Examples

# Single dimension
PartInv(propsel = .10,
        weights_item = c(1, 0.9, 0.8, 1),
        weights_latent = 1,
        alpha_r = 0.5,
        alpha_f = 0,
        psi_r = 1,
        lambda_r = c(.3, .5, .9, .7),
        nu_r = c(.225, .025, .010, .240),
        nu_f = c(.225, -.05, .240, -.025),
        Theta_r = diag(.96, 4),
        labels = c("Female", "Male"))
# multiple dimensions
lambda_matrix <- matrix(0,nrow = 5, ncol = 2)
lambda_matrix[1:2, 1] <- c(.322, .655)
lambda_matrix[3:5, 2] <- c(.398, .745, .543)
PartInv(propsel = .05,
        weights_latent = c(0.5, 0.5),
        alpha_r = c(0, 0),
        alpha_f = c(-0.3, 0.1),
        psi_r = matrix(c(1, 0.5, 0.5, 1), nrow = 2),
        lambda_r = lambda_matrix,
        nu_r = c(.225, .025, .010, .240, .125),
        nu_f = c(.225, -.05, .240, -.025, .125),
        Theta_r = diag(1, 5),
        Theta_f = c(1, .95, .80, .75, 1))
PartInvMulti_we(propsel = .10,
                weights_item = c(1/3, 1/3, 1/3, 1/3),
                weights_latent = 1,
                alpha_r = 0.5,
                alpha_f = 0,
                psi_r = 1,
                lambda_r = c(.3, .5, .9, .7),
                nu_r = c(.225, .025, .010, .240),
                nu_f = c(.225, -.05, .240, -.025),
                Theta_r = diag(.96, 4),
                labels = c("female", "male"))

claycantrell/PartInvShinyUI documentation built on March 29, 2022, 9:49 a.m.