lmhelprs (version 0.4.0)

lm_list_to_partable: Convert a 'lm_list' Object To a Parameter Table

Description

Convert the output of many_lm() to a lavann-style parameter table.

Usage

lm_list_to_partable(
  object,
  keep_intercepts = FALSE,
  vcov_args = list(),
  pvalue_fun = NULL,
  rsquare = FALSE
)

Value

A data frame object with columns such as lhs, op, rhs, and est, major columns of the output of lavaan::parameterTable() necessary for plotting the model using semPlot::semPaths().

Arguments

object

The output of many_lm() or manymome::lm2list().

keep_intercepts

Logical. If TRUE, the intercepts of the regression models and the means of the "pure" predictors (variables not being the outcome variables of any of the regression models) are kept in the parameter table. If FALSE, the default, all intercepts and means will be removed.

vcov_args

A named list of arguments to be passed to stats::vcov() when computing the standard errors of the regression coefficients. Default is list(), an empty list.

pvalue_fun

The function to be used to compute the p-values of regression coefficients. Ignored for now. Included for adding this feature in the future.

rsquare

Logical. Whether R-squares will be included in the output, with r2 as the operator in the column op. Default is FALSE. Not be included by default because semPlot::semPaths() will draw the R-squares over the residual variances.

Author

Shu Fai Cheung https://orcid.org/0000-0002-9871-9448

Details

This function convert a a lit of lm objects, such as the output of many_lm() or manymome::lm2list(), to a table of parameter estimates similar to the output of lavaan::parameterTable.

The output is designed to be used by semPlot::semPaths() and so contains only information necessary for the plot.

The output of stats::lm() is already supported by semPlot::semPaths(), and it can also combine a list of regression models into on single plot. However, it will convert interaction terms to knots. Moreover, if two interaction terms in two different models share the a variable, it will be incorrectly combined to become a single knot (Version 1.1.6). Therefore, this function was developed to let users to draw the model as if it were a path model in structural equation modeling.

See Also

many_lm() and manymome::lm2list().

Examples

data(data_test1)
mod <- "x3 ~ x2*x1
        x4 ~ x3
        x5 ~ x4 + x3"
out <- many_lm(mod, data_test1)
out_ptable <- lm_list_to_partable(out)
out_ptable

m <- matrix(c("x1", "x2", "x2:x1", NA, "x3", NA, "x4", NA, NA, NA, "x5", NA),
            nrow = 3, ncol = 4)
m

# The output can be used by semPlot::semPaths()

if (requireNamespace("semPlot", quietly = TRUE)) {
  library(semPlot)
  p <- semPaths(out_ptable,
                what = "paths",
                whatLabels = "est",
                nCharNodes = 0,
                style = "ram",
                layout = m,
                exoCov = FALSE,
                DoNotPlot = TRUE)
  plot(p)

  # If it is desired to use knots to
  # denote interaction terms, then,
  # the output of many_lm() can be used
  # directly.

  m2 <- matrix(c("x1", NA, "x2", NA, "x3", NA, "x4", NA, NA, NA, "x5", NA),
            nrow = 3, ncol = 4)
  p2 <- semPaths(out,
                 what = "paths",
                 whatLabels = "est",
                 nCharNodes = 0,
                 style = "ram",
                 layout = m2,
                 exoCov = FALSE,
                 intercepts = FALSE,
                 DoNotPlot = TRUE)
  plot(p2)

  # This illustrates the problem with using
  # the list of lm-outputs directly when
  # a variable is involved in the interaction terms
  # of two or more models.

  m3 <- matrix(c("x2",   NA, "x1",   NA, "x3",
                   NA,   NA,   NA,   NA,   NA,
                   NA, "x4",   NA, "x5",   NA),
            nrow = 5, ncol = 3)
  mod3 <- "x4 ~ x2*x1
           x5 ~ x3*x1"
  out3 <- many_lm(mod3, data_test1)
  p3 <- semPaths(out3,
                 what = "paths",
                 whatLabels = "est",
                 nCharNodes = 0,
                 style = "ram",
                 layout = m3,
                 exoCov = FALSE,
                 intercepts = FALSE,
                 DoNotPlot = TRUE)
  plot(p3)

}