Dear statisticians, I conducted regression containing main effects and interaction effects. Main effects had a signifecant effect on Y, However interactions of them were not signifecant. So what does it mean? and what is the evidence of that?
You judged it non-significant. That is your judgement, How can't you know what it means?
If you's say "the p-value for the interaction was 0.23 - what does it mean?" then I would answer: It means that the probability to get a larger F-value under the assumption of purely additive effects (i.e.: no interaction) is 0.23. This is what a statistician can say. What this means to you is something you have to decide. If you find this probability large, then you might be of the opinion that the interaction term in the model is not required to describe the main features in your data. If your data is generated in a designed experiment to specifically test the interaction, then you might conclude that the data you gathered was not sufficient to demonstrate the interaction with sufficient confidence. But this is all a lot of subject matter and not statistics.
A (significant, meaningful) interaction means that the effect of one factor on the y-variable is not consistent across the levels of the other factor.
Imagine a two-way anova with one factor being sex (boys and girls) and the other being age as a categorical variable ( age 12, 14, 18 ). And we are measuring height.
Maybe there is a significant effect of sex: boys are taller than girls overall. And maybe there is a significant effect of age: older children are taller.
A significant interaction would mean that the effect of sex was not consistent across the levels of age (or vice-versa).
One way this could happen is if boys are taller than girls at every age group, but that the difference between boys and girls is less at age 12.
Another way this could happen is if the girls are actually taller than the boys at age 12.