MANCOVA interaction question

DrLes

New Member
#1
Hi,
I have what is probably a simple (or self-explanatory question). I used a MANOVA with three DV's, two IV's and about 6 covariates. I found significant effects where I expected them, but I also expected to see a significant interaction between my two IVs. However, when I plot them, it LOOKS like an interaction...the lines are crossing over one another for two of the three DVs. My question is, if I report the non-significant interaction, can I still display the line graphs? If so, how would I address this? What I want to say is, "The non-significant interaction provides further evidence for the independent predictive nature of each variable". Is this true? I'm confusing myself more by the minute. Can anyone offer some help/insight? THANK YOU!
 

trinker

ggplot2orBust
#2
The interaction plots you refer to are generally used to determine where the means differ as the two variables interact (sorry I'm no statistician so you get the description according to my understanding). If there is no significant difference in interactions then there's no need to refer to where differences lie. This is albeit to reporting a strength of effect measure for non-significant effects. In journals paper is valuable and mentioning the difference found in the sample, not attributable to the population is in itself a waste of space let alone a graph. The exception to this may be if your study is a pilot study. This is my take on it, others may weigh in differently.

Edit: caveat about the strength of effect comment for non sig. tests- some journals require this though I believe this to inappropriate.
 
Last edited: