Tabulating results from structural equation models"

In tidySEM: Tidy Structural Equation Modeling

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)
run_mplus <- FALSE

library(tidySEM)
library(lavaan)
library(MplusAutomation)

tidySEM tabulates the results of different types of models in the same uniform way. This facilitates parsing the output into Tables and Figures for publication. The function to tabulate output is table_results(). The function to tabulate fit indices is table_fit().

Let's use a classic lavaan tutorial example for a multiple group model, using the HolzingerSwineford1939 data. The tidySEM package has a function measurement() to generate measurement models automatically. It guesses which latent variable an observed variable belongs to by splitting the names (by default, at the last _ symbol), so it helps to rename the variables:

df <- HolzingerSwineford1939
names(df)[7:15] <- paste0(rep(c("vis", "tex", "spe"), each = 3), "_", rep(1:3, 3))
df |>
  subset(select = c("school", "vis_1", "vis_2", "vis_3", "tex_1", "tex_2", "tex_3", "spe_1", 
"spe_2", "spe_3")) -> df

df <- HolzingerSwineford1939
names(df)[7:15] <- paste0(rep(c("vis", "tex", "spe"), each = 3), "_", rep(1:3, 3))
subset(df, select = c("school", "vis_1", "vis_2", "vis_3", "tex_1", "tex_2", "tex_3", "spe_1", 
"spe_2", "spe_3")) -> df

Now, let's construct the model.

df |>
  tidy_sem() |>
  measurement() -> model

model <- measurement(tidy_sem(df))

Output from lavaan

Now, let's run the model in lavaan. You can either use lavaan to run it,

model |>
  estimate_lavaan() -> fit_lav

estimate_lavaan(model) -> fit_lav

The results can be tabulated using table_results():

table_results(fit_lav)

table_fit(fit_lav)

Output from OpenMx

Now, we'll reproduce the same analysis in 'OpenMx'. First, we run the model:

model |>
  estimate_mx() -> fit_mx
table_results(fit_mx)
table_fit(fit_mx)

estimate_mx(model) -> fit_mx
table_results(fit_mx)
table_fit(fit_mx)

Output from Mplus

Now, we'll reproduce the same analysis in 'Mplus'. To illustrate the fact that tidySEM is compatible with existing solutions, we will specify the syntax for this example manually, using the package MplusAutomation. This code will only work on your machine if you have Mplus installed and R can find it. First, we run the model:

fit_mplus <- mplusModeler(mplusObject(VARIABLE = "grouping IS school (1 = GW 2 = Pas);",
                                MODEL = c("visual BY vis_1 vis_2 vis_3;",
                                          "textual BY tex_1 tex_2 tex_3;",
                                          "speed BY spe_1 spe_2 spe_3;"),
                                usevariables = names(df),
                                rdata = df),
                    modelout = "example.inp",
                    run = 1L)
table_results(fit_mplus)

# fit <- mplusModeler(mplusObject(VARIABLE = "grouping IS school (1 = GW 2 = Pas);",
#                                 MODEL = c("visual BY x1 x2 x3;",
#                                           "textual BY x4 x5 x6;",
#                                           "speed BY x7 x8 x9;"),
#                                 usevariables = c(paste0("x", 1:9), "school"),
#                                 rdata = HolzingerSwineford1939),
#                     modelout = "example.inp",
#                     run = 1L)
# file.remove(list.files(pattern = "^example.+(inp|out|dat)$"))
#dput(fit$results$parameters)
#dput(fit, file = "mplusfit.R")

# Read the results
fit_mplus <- source("mplusfit.R")
fit_mplus <- fit_mplus$value

table_results(fit_mplus)
table_fit(fit_mplus)

tidySEM documentation built on Oct. 25, 2023, 1:06 a.m.