Plot Factor Loadings for lavaan Models

Creates a dot plot of standardized factor loadings for a fitted lavaan model, including CFA and SEM. The plot uses only the measurement model portion (=~ parameters), along with optional confidence intervals for each item. If the model does not contain =~ parameters, the function errors with a clear message.

Usage

plot_factor_loadings(
  fit,
  sort = TRUE,
  group_by = TRUE,
  standardized = TRUE,
  ci = TRUE,
  autofit = TRUE,
  ci_bounds = c("extend", "arrow"),
  facet_by = c("none", "group", "level", "group_level"),
  verbose = TRUE,
  ...
)

Arguments

fit

A fitted lavaan model object (including CFA and SEM).

sort

Logical; if TRUE, sorts items by loading size. Defaults to TRUE.

group_by

Logical; if TRUE and the model has multiple factors, groups the items by factor. Defaults to TRUE.

standardized

Logical; if TRUE, uses standardized loadings. Defaults to TRUE.

ci

Logical; if TRUE, includes confidence intervals. Defaults to TRUE. Also accepts CI via ... for backward compatibility.

autofit

Logical; if TRUE, computes and applies x-axis limits via coord_cartesian() based on loadings (and CIs when available). If all values are non-negative, the lower limit is set to 0; if all values are non-positive, the upper limit is set to 0. For standardized loadings within [-1, 1], the limits are extended to include the nearest boundary to keep the standardized scale visible. If FALSE, the x-axis limits are not modified and ggplot2 determines the range.

ci_bounds

Controls how confidence intervals are handled for standardized loadings when ci = TRUE and autofit = TRUE. "extend" draws full CIs (axis may extend beyond [-1, 1]); "arrow" constrains the x-axis to [-1, 1] (or [0, 1]/[-1, 0] when all values are non-negative/non-positive), clips CIs to that range, and adds arrows to indicate off-scale intervals. If any standardized point estimate is outside [-1, 1], "arrow" falls back to "extend".

facet_by

Character string controlling faceting for SEM contexts with grouping metadata. Options:

• "none" (default): no faceting.

• "group": facet by lavaan group.

• "level": facet by lavaan level (multilevel models).

• "group_level": facet by combined group and level. If requested metadata is unavailable, a message is shown when verbose = TRUE and plotting continues without faceting.

verbose

Logical; if TRUE, prints informational messages and warnings (for example, when the model did not converge). Defaults to TRUE. When FALSE, non-fatal messages and warnings from internal lavaan extraction calls are suppressed; errors are still raised.

...

Additional arguments passed to ggplot2::ggplot.

Value

A ggplot object if ggplot2 is installed, otherwise an error message.

See also

• plot-methods for an overview of plotting in the package.

• plot.lavaan() for more lavaan object plots.

Examples

if (requireNamespace("lavaan", quietly = TRUE) &&
    requireNamespace("ggplot2", quietly = TRUE)) {
  library(lavaan)
  library(psymetrics)
  hs_model <- 'visual  =~ x1 + x2 + x3
              textual =~ x4 + x5 + x6
              speed   =~ x7 + x8 + x9'
  fit <- cfa(hs_model, data = HolzingerSwineford1939, estimator = "MLR")
  plot_factor_loadings(fit)
  sem_model <- 'visual  =~ x1 + x2 + x3
                textual =~ x4 + x5 + x6
                speed   =~ x7 + x8 + x9
                textual ~ visual
                speed ~ textual'
  fit_sem <- sem(sem_model, data = HolzingerSwineford1939, group = "school")
  plot_factor_loadings(fit_sem, facet_by = "group")
} else {
  message("Please install 'lavaan' and 'ggplot2' to run this example.")
}