(Bayesian) Process Model with Control Variables

Comments

• EJ

  April 18

  Hi Mila,

  I have attended the experts to this! Sorry about the huge delay in responding

  EJ
• JohnnyB

  April 20

  Hi @Mila_Marinova,

  I think it depends on what you mean by "controlling for", and which pathways you draw depends on the association structure you have in mind between your predictors (independent variables and covariates/controls).

  • Do you want to test whether your control variable is affecting the mediation effect? Then you could add a moderation pathway between that control and whichever path you think it's moderating.
  • Do you simply want to add it as a predictor? Then you could draw direct paths from the control variable to Y, to the mediator(s), and optionally to X - based on which relationships you believe exist. Note that if your control variable affects the mediator but you omit that path, the estimate of the X -> M relationship will be biased.

  Having the model builder in jaspProcess encourages you to think more carefully about what specific model you are testing. It's a bit more effort but hopefully leads to more careful modeling, rather than picking a number from a catalogue and forcing your observed variables into that pattern. Additionally, your theory (and research design) determines which paths are plausible, not the software: JASP will fit any model you draw, but a statistically significant path is not evidence of a causal connection.

  Kind regards,

  Johnny