DEMO: SEMinR - cSEM comparison (response to comment on Reddit)

seminr-csem-compare.R

# Code below demonstrates similar/same results from SEMinR and cSEM for given model
# Needed changing default value of .disattenuate for cSEM
# Response to: https://www.reddit.com/r/rstats/comments/1fy42vy/structural_equation_model_results_differ_when/

library(seminr)
library(cSEM)
nck <- corp_rep_data
data <- MASS::mvrnorm(1000,mu = rep(0,37),Sigma =diag(1,37))
data <- cbind(mobi, corp_rep_data[1:250,3:17])
colnames(data) <- c(multi_items("A", 1:3),
                    multi_items("B", 1:3),
                    "K1",
                    multi_items("C", 1:5),
                    multi_items("D", 1:4),
                    multi_items("E", 1:3),
                    multi_items("F", 1:2),
                    multi_items("G", 1:3),
                    multi_items("H", 1:3),
                    multi_items("I", 1),
                    multi_items("J", 1:9)
                    )

### Seminr ----
mm <- constructs(
  composite("Construct_A", multi_items("A", 1:3), weights = mode_A),
  composite("Construct_B", multi_items("B", 1:3), weights = mode_A),
  composite("Construct_D", multi_items("D", 1:3), weights = mode_A),
  composite("Construct_E", multi_items("E", 1:3), weights = mode_B),
  composite("Construct_F", multi_items("F", 1:2), weights = mode_B),
  composite("Construct_G", multi_items("G", 1:3), weights = mode_B),
  composite("Construct_H", multi_items("H", 1:3), weights = mode_B),
  composite("Construct_I", single_item("I1")),
  composite("Construct_J", c(multi_items("J", 1:9), "K1"), weights = mode_B)
)

sm <- relationships(
  paths(from = c("Construct_I", "Construct_H"),
        to = c("Construct_A")),
  paths(from = c("Construct_A",
                 "Construct_B",
                 "Construct_E",
                 "Construct_F",
                 "Construct_G"),
        to = c("Construct_D")),
  paths(from = c("Construct_E", 
                 "Construct_D"),
        to = c("Construct_J")
))

model <- estimate_pls(
  data = data,
  measurement_model = mm,
  structural_model = sm,
  inner_weights = path_factorial,
  missing = mean_replacement,
  missing_value = "-99"
)

# summary_model <- summary(model)
plot(model)
# summary_model
# 
# summary_model$iterations
# 
# summary_model$validity$vif_items
# 
# summary_model$paths
# 
# summary_model$reliability


# CSEM ----
model_csem <- "
Construct_A =~ A1 + A2 + A3 # Mode A (reflective by default)
Construct_B =~ B1 + B2 + B3 # Mode A (reflective by default)
Construct_D =~ D1 + D2 + D3

Construct_E <~ E1 + E2 + E3 # Mode B (formative)
Construct_F <~ F1 + F2  # Mode B (formative)
Construct_G <~ G1 + G2 + G3 # Mode B (formative)
Construct_H <~ H1 + H2 + H3# Mode B (formative)
Construct_I =~ I1 
Construct_J <~ J1 + J2 + J3 + J4 + J5 + J6 + J7 + J8 + J9 + K1 # Mode B (formative)


Construct_A ~ Construct_I + Construct_H
Construct_D ~ Construct_A + Construct_B+ Construct_E + Construct_F + Construct_G
Construct_J ~ Construct_D + Construct_E
"


model_run <- csem(

  .model = model_csem,
  .disattenuate = FALSE,
  .data = data,

)
model$path_coef
model_run$Estimates$Path_estimates
summary(model_run)