plsem function

S3 interface for semi-confirmatory SEM via PL

plsem() is an S3 interface for obaining a fitted lslx object.

plsem(
  model,
  data,
  penalty_method = "mcp",
  lambda_grid = "default",
  delta_grid = "default",
  numeric_variable,
  ordered_variable,
  weight_variable,
  auxiliary_variable,
  group_variable,
  reference_group,
  sample_cov,
  sample_mean,
  sample_size,
  sample_moment_acov,
  verbose = TRUE,
  ...
)

Arguments

• model: A character with length one to represent the model specification.
• data: A data.frame of raw data. It must contains variables specified in model (and possibly the variables specified by group_variable and weight_variable).
• penalty_method: A character to specify the penalty method. The current version supports "none", "lasso", "ridge", "elastic", and "mcp".
• lambda_grid: A non-negative numeric to specify penalty levels for both "lasso" and "mcp". If it is set as "default", its value will be generated automatically based on the variable scales.
• delta_grid: A non-negative numeric to specify the convexity level for "mcp". If it is set as "default", its value will be generated automatically based on the variable scales.
• numeric_variable: A character to specify which response variables should be transfromed into numeric.
• ordered_variable: A character to specify which response variables should be transfromed into ordered.
• weight_variable: A character with length one to specify what variable is used for sampling weight.
• auxiliary_variable: A character to specify what variable(s) is used as auxiliary variable(s) for estimating saturated moments when missing data presents and two-step method is implemented. Auxiliary variable(s) must be numeric. If any categorical auxiliary is considered, please transform it into dummy variables before initialization.
• group_variable: A character with length one to specify what variable is used for labeling group.
• reference_group: A character with length one to specify which group is set as reference.
• sample_cov: A numeric matrix (single group case) or a list of numeric matrix (multi-group case) to represent sample covariance matrixs. It must have row and column names that match the variable names specified in model.
• sample_mean: A numeric (single group case) or a list of numeric (multi-group case) to represent sample mean vectors.
• sample_size: A numeric (single group case) with length one or a list of numeric (multi-group case) to represent the sample sizes.
• sample_moment_acov: A numeric matrix (single group case) or a list of numeric matrix (multi-group case) to represent asymptotic covariance for moments.
• verbose: A logical to specify whether messages made by lslx should be printed.
• ...: Other arguments. For details, please see the documentation of lslx.

Returns

A fitted lslx object

Examples

## EXAMPLE: Semi-Confirmatory Factor Analysis with lavaan Style ##
# specify a factor analysis model with lavaan style
model_fa <- "visual  =~ x1 + x2 + x3
             textual =~ x4 + x5 + x6
             speed   =~ x7 + x8 + x9
             pen() * visual  =~ x4 + x5 + x6 + x7 + x8 + x9
             pen() * textual =~ x1 + x2 + x3 + x7 + x8 + x9
             pen() * speed   =~ x1 + x2 + x3 + x4 + x5 + x6
             visual  ~~ 1 * visual
             textual ~~ 1 * textual
             speed   ~~ 1 * speed"
             
# fit with mcp under specified penalty levels and convexity levels
lslx_fa <- plsem(model = model_fa, 
                 data = lavaan::HolzingerSwineford1939,
                 penalty_method = "mcp", 
                 lambda_grid = seq(.02, .60, .02), 
                 delta_grid = c(1.5, 3.0, Inf))

# summarize fitting result under the penalty level selected by 'bic'
summary(lslx_fa, selector = "bic")