fanc - Penalized Likelihood Factor Analysis via Nonconvex Penalty
Computes the penalized maximum likelihood estimates of
factor loadings and unique variances for various tuning
parameters. The pathwise coordinate descent along with EM
algorithm is used. This package also includes a graphical tool
which outputs path diagrams, heatmaps, goodness-of-fit indices
and model selection criteria for each regularization parameter
(Yamamoto, M., Hirose, K. and Nagata, H., 2017
<doi:10.1007/s41237-016-0007-3>). The user can change the
regularization parameter interactively with a built-in
self-contained HTML viewer (no additional packages required),
which is helpful to find a suitable value of regularization
parameter. As a penalty, we can choose either the minimax
concave penalty (Hirose, K. and Yamamoto, M., 2015
<doi:10.1007/s11222-014-9458-0>; Hirose, K. and Yamamoto, M.,
2014 <doi:10.1016/j.csda.2014.05.011>) or the product-based
elastic net penalty (Hirose, K. and Terada, Y., 2023
<doi:10.1007/s11336-022-09868-4>).
