latcontrol {latcontrol}

Evaluation of the equivalence of the model-implied matrices of structural equation models with and without control variables

Description

Evaluation of the model-implied variance-covariance matrices of two structural equation models that only differ by the inclusion versus exclusion of one or more control variable(s). Both models need to be fitted with the R package lavaan (Rosseel, 2012) <doi:10.18637/jss.v048.i02>. The derivation of the methodology employed in this package can be obtained from Blötner (2023) <doi:10.31234/osf.io/dy79z>.

Usage

latcontrol(object_with, object_without, type = c("simple", "complex"))

Arguments

object_with

Fit object from the 'lavaan' package (Rosseel, 2012 <doi:10.18637/jss.v048.i02>) with the control variable(s).

object_without

Fit object from the 'lavaan' package (Rosseel, 2012 <doi:10.18637/jss.v048.i02>) without the control variable(s).

type

Optional. Specifies whether a single-level structural equation model or a multilevel structural equation model is entered (DEFAULT = "simple").

Details

The latcontrol function itself was derived from the discrepancy function from confirmatory factor analysis and structural equation models. In analogy to the latter latent model classes, the function provides a chi-square-based index of discrepancy, model degrees of freedom, a p-value, and derivatives of common descriptive model fit indices (i.e., Root Mean Square Error of Approximation and Square Root Mean Residual).

Value

X2

Chi-square value, reflecting the difference between the two matrices.

df

Degrees of freedom of the Chi-square statistic.

p_value

Corresponding p-value of the Chi-square statistic with the stated degrees of freedom.

rmsea

Adapted version of the Root Mean Square Error of Approximation to evaluate whether the two matrices differ.

srmr

Adapted version of the Square Root Mean Residual to evaluate whether the two matrices differ.

Author(s)

Christian Blötner [e-mail]

References

Blötner, C. (2023). latcontrol: Evaluation of the role of control variables in structural equation models. PsyArXiv. https://doi.org/10.31234/osf.io/dy79z

Rosseel, Y. (2012). lavaan: An R package for structural equation modeling. Journal of Statistical Software, 48(2), 1–36. https://doi.org/10.18637/jss.v048.i02

Examples

data <- data.frame(i1 = rnorm(100),
                   i2 = rnorm(100),
                   i3 = rnorm(100),
                   i4 = rnorm(100),
                   i5 = rnorm(100),
                   i6 = rnorm(100),
                   i7 = rnorm(100),
                   i8 = rnorm(100),
                   i9 = rnorm(100),
                   i10 = rnorm(100),
                   i11 = rnorm(100),
                   i12 = rnorm(100))

m_with <- 'IV =~ i1 + i2 + i3 + i4
           DV =~ i5 + i6 + i7 + i8
           CV =~ i9 + i10 + i11 + i12

           DV ~ IV + CV
           IV ~ CV'
m_without <- 'IV =~ i1 + i2 + i3 + i4
              DV =~ i5 + i6 + i7 + i8

              DV ~ IV'

fit_with <- sem(m_with, data = data)
fit_without <- sem(m_without, data = data)

latcontrol(fit_with, fit_without)