3  Approaches

There are a number of approaches for estimating interaction effects in SEM. In modsem(), the method = “method” argument allows you to choose which to use.

• “ca” = constrained approach (Algina & Moulder, 2001)
  • not recommended for cases where there is a main effect between variables in the interaction term, unless you know what you are doing.
  • e.g.,
  • use removeFromParTable = “X ~~ Z”
  • and addToParTable = “new formula for covariance, with label Cov_X_Z”
• “uca” = unconstrained approach (Marsh, 2004)
  • not recommended for cases where there is a main effect between variables in the interaction term, unless you know what you are doing.
• “rca” = residual centering approach (Little et al., 2006)
  • default
• “dblcent” = double centering approach (Marsh., 2013)
• “pind” = basic product indicator approach (not recommended)
• “lms” = The Latent Moderated Structural equations approach through the nlsem package
  • note: there can not be an interaction between two endogenous variables.
• “qml” = The Quasi Maximum Likelihood approach.
  • note: can only be done if you have a single endogenous (dependent) variable.
• “mplus”
  • estimates model through Mplus, if it is installed

modsem(m1, oneInt, "ca")
modsem(m1, oneInt, "uca")
modsem(m1, oneInt, "rca")
modsem(m1, oneInt, "dblcent")
modsem(m1, oneInt, "mplus")
modsem(m1, oneInt, "lms")
modsem(m1, oneInt, "qml")