Plotting latent interaction

60 views
Skip to first unread message

Janitor

unread,
Sep 30, 2020, 11:45:59 AM9/30/20
to lavaan
Hi,
I'm trying to plot a latent interaction which I tested using the probe2WayMC-function. I have encountered several problems: 

1. If I use the plotProbe-function, the dependent variable is on the y-axis and the independent on the x-axis, which is rather counter-intuitiv. Is there a way to turn them around? OR is there another package I could use to produce a paper-ready-graph? I already read about the quantpsyc-website but was not able to retrieve coef(fit) and vcov(fit), as suggested by prof. jorgensen in another conversation, for my s4-class-lavaan-object.

2.  I used age (4-16 years) as a continous variable for my latent-interaction-terms. Now I'm not sure how to set the valProbe to best illustrate the effect?

I would really appreciate any tipps for problem solving.

Alex Schoemann

unread,
Sep 30, 2020, 3:40:36 PM9/30/20
to lavaan
To answer your questions:

1. I'm not sure why you want to have the outcome on the x axis and the predictor on the y axis. Almost every plot of two variables includes the outcome (Y variable) on the y axis and predictor (X variable) on the x axis. If you really want to change things, the output from probe2WayMC should give you the simple intercepts and slopes, with an intercept and slope you can then plot a line in any number of plotting packages/software.

2. Pick the values that are meaningful. There's no rule that says which values you need to pick. With a latent moderator the default is +-1 SD since the metric of the latent variable isn't meaningful. With age you would pick the values that are most relevant.

Reply all
Reply to author
Forward
0 new messages