Hi Zahra,

I asume, your interaction is between quantitative X nad binary moderator D.

Report the usual regression table with the coefficients of the first order effects of X and D and the product term of XD.

If the product term is signficiant (supporting your moderator hypothesis), generate a simple slope plot that shows the effect of X for D=0 and for D=1. There are excel sheets out there (google for Jeremy Dawson) or if you use R, ther is the nice interactions package:

install.packages("interactions")

library(interactions)

interact_plot(

mod_reg, #name of the model object

pred = X,

modx = D,

grid.values = "1SD",

x.label = "X",

y.label = "name_of_dv",

legend.main = "D" #displays the variable D for the legend

)

Best,

Holger

The main question is what should I report? For example, in my research, the interaction is not significant in general, but when I use the condition, it is significant in the first group. I don't know which should be reported: the general interaction or just the conditional interaction?

Hi Zahra,

so, your interaction (=product term) is not significant but one group has a significant effect and the other does not? This is not usual. This happens when one effect is slightly above the "significance barrier" and the other isn't. However, the difference between both is not significant. The last issue is relevant.

In a structural equation modeling context (perhaps this could be attractive for you) we would test the equality constraint for both parameters and test whether the model is significantly worsend. As this test would be passed in your case, we would the commonly estimated coefficient as the best candidate for both groups.

HTH

Holger

Dear Holger,

You can look at this and see the result. I use the PROCESS for dichotomous moderator (vaccinated /not vaccinated). Holger Steinmetz

Hi Zahra,

yes, as I said. By the way: You should dummy code the vaccination-variable (0/1).

Best,

Holger

Thank you, I am very grateful for your response.