Interaction Plot error: The lines do not meet

Question

Hello I wanted to develop an alternative regression equation just to prove that my first regression equation is a better fit.

I want to show the possibility that my independent variables my be interacting with each other. To show that I wanted to plot an interaction graph.

I will first write the multiple regression equation and then the interaction plot and also the resulting plot.

> interact_plot(Poly, pred = x2, modx = x1)
Warning message:
-1.13290847281745 is outside the observed range of x1 

What does this warning message mean?

And I do not really know how to interpret this? Does this mean they are no interacting? Do they have to meet at some point?

Thanks for your help, really in need.

Answer 1

This question is better suited for Cross Validated, but the answers are pretty straightforward:

1. What does this warning message mean?

From the arguments in interact_plot() it looks like you're using jtools. The jtools::interact_plot documentation notes that the modxvals argument can be used to specify the range of moderator values.

If modxvals is left to the default (NULL), then:

the customary +/- 1 standard deviation from the mean as well as the mean itself are used for continuous moderators.

But there's no guarantee that your actual observed values include values up to +/- 1 SD from the sample mean.

The warning you're seeing was added as a heads up that the plot is using moderator values that don't appear in your data, but which do fall within the +/- 1 SD default range. You most likely don't need to worry about this.

2. Do they have to meet at some point?

An interaction effect does not need to have moderator slopes "cross" in the plot of the data, but you will see that the angles of the slopes would cross at some point, if the plot extended far enough along the x-axis. In the plot you posted, imagine if the x values continued off the left end of the graph - pretty soon, the slopes you see would, in fact, overlap.

You can usually rely on this rule of thumb for interpreting visualizations of interaction effects. (But better to check the coefficients in your model, to make sure.)