Exploratory factor analysis on multiple data imputations

EFA_POOLED() has been superseded by efa_mi(), which is the recommended interface going forward. It remains available and unchanged so existing code keeps working.

Usage

EFA_POOLED(
  data_list,
  p = 0.05,
  target_method = c("first_target", "consensus"),
  align_unrotated = c("signed_tucker_congruence", "none", "procrustes"),
  fit_pool_method = c("D2"),
  consensus_args = list(),
  procrustes_args = list(),
  rmsea_ci_level = 0.9,
  rmsr_upper = TRUE,
  ...
)

Arguments

data_list

A list of length \(m\), where \(m\) is the number of imputations. Each list element is a data frame or matrix of raw data, or a correlation matrix. See argument x in efa_fit(). A mids object from mice must be converted first, with mice::complete(x, "all").

p

Numeric in \((0, 1)\). One minus the confidence level for the pooled confidence intervals, whichever se method produced them ("information", "np-boot", or "sandwich"). For example, p = .05 gives 95% intervals.

target_method

Character. How rotated solutions are aligned across imputations before pooling: "first_target" (the default) aligns every imputation to the first imputation's rotated solution, while "consensus" refines a centroid target by Generalized Procrustes Analysis, started from the medoid imputation so that the pooled rotated solution does not depend on the order of data_list (orthogonal rotations only). See Aligning solutions across imputations in Details.

align_unrotated

Character. How unrotated loadings are aligned before pooling: "signed_tucker_congruence" (the default; sign/permutation via Tucker congruence, anchored on the medoid imputation and returned in the extraction's canonical gauge), "procrustes" (orthogonal Procrustes to the first imputation), or "none". See Aligning solutions across imputations in Details.

fit_pool_method

Character. Only "D2" is implemented for pooling chi-square-type fit. If no chi-square is available, only residual-based fit and descriptive quantities are returned. See Pooling the model chi-square and fit indices in Details.

consensus_args

List of additional arguments controlling the GPA-consensus iteration when target_method = "consensus". Recognised tuning parameters include the convergence tolerances tol and loss_tol, the iteration bounds min_iter and max_iter, the target-update damping alpha, the multi-start controls multi_start and starts, and start, which overrides the medoid imputation the iteration is otherwise started from.

procrustes_args

List of efa_procrustes() algorithm controls for fixed-target alignment, for example oblique_maxit or oblique_random_starts. The loadings A, the alignment Target, the rotation family, and the cross-product S are derived from the imputations and cannot be set here.

rmsea_ci_level

Numeric. Confidence level for the RMSEA CI.

rmsr_upper

Deprecated and ignored. efa_mi() now always computes RMSR the same way, from the unique off-diagonal residuals; SRMR is reported alongside it. Supplying it to efa_mi() signals a deprecation warning; the superseded EFA_POOLED() accepts it silently.

...

Additional arguments passed to efa_mi() and from there to efa_fit(), in the flat form of the former interface (e.g. method, rotation, type, max_iter, or the extras of the selected rotation such as gam). Names that the former interface could not use are ignored.

Value

The value of efa_mi(), normally a list of class c("efa_mi", "EFA_POOLED", "efa", "EFA"); see there for the components.

See also

efa_mi()