Interaction term significant but in the plot the lines don't intersect

Question

I tested the moderating effect of the variable X' on the relation between X and Y. The results were positive and significant.

However, when I plotted the results, I found out the two lines become very, very close at the higher levels, but do not "cut each other".

I wonder if these results make sense... I'm trying to clarify if X' in fact moderates the relation between X-Y (the plot is confusing me).

Thanks!

There's nothing about 'interaction' that implies 'a plot of the line segments joining the means must intersect', only that the corresponding population line segments won't be parallel. (It would help if you plotted the data, or gave more information on the variables, or something similar.) — Glen_b, Nov 28 '13 at 22:54

score 8 · Answer 1 · answered Nov 28 '13 at 13:16

8

In order for an interaction to exist the slopes of the two lines should be different. It is not necessary for the lines to cross within the range of the data.

answered Nov 28 '13 at 13:16

Maarten Buis

21,005

@peter-flom this time you were a couple of seconds faster... – Maarten Buis Nov 28 '13 at 13:17

score 6 · Answer 2 · answered Nov 28 '13 at 13:16

6

There is no need for the lines to intersect for there to be an interaction. All that is needed is that the lines are not parallel.

Note also that recent work on moderation gets away from the older "moderates", "does not moderate" dichotomy and more into "how much moderation".

answered Nov 28 '13 at 13:16

Peter Flom

119,535
36
175
383

1

As with @Maarten Buis's answer, pedants will want to insert "within the range of the data" in your first sentence. – Nick Cox Nov 28 '13 at 15:38

Interaction term significant but in the plot the lines don't intersect

2 Answers2