I'm examining a moderated-moderation model (IV = marital conflict; Moderator 1 = emotion regulation; Moderator 2 = Marital Status; DV = Marital Satisfaction). I used SEM to run two separate models: a model of never divorced couples and a model of divorced couples. I found that emotion regulation moderated the negative relationship between marital conflict and marital satisfaction for never divorced couples but not for divorced couples. To test whether the observed interaction effect held across both couple types, I constrained the interaction paths to be equal. Although the interaction term was significant for never divorced couples and not for divorced couples, the chi-square difference test revealed that there was not a difference between the interaction estimates. Does this mean that the significant interaction emerged due to chance? Or is there another explanation? Thank you for your time