Can anyone help me understand why the independent variable was rendered insignificant when M1 variable was entered into the model but then regained significance when I included M2 (which is correlated with M 1)?
x --> y = significant x --> m1 --> y = x lost significance (indirect effect M significant) x --> m1 --> m2-->y = x regained significance, M1 and M2 also signifiant
Could it have to do with the correlation between M1 and M2? I looked into the possibility of suppression, but everything I'm reading suggests that suppression assumes that the "suppressor" variable is not associated with Y, but all of my variables are related. In fact, M2 is the strongest predictor of Y. All variables are dichotomous. Thank you
I mean that x-y was sig in the first model, but then lost significance in the second model (suggesting that the effect of x on y was mediated through m). So, I'm wondering why x would become significant again after adding a second mediator in the 3rd/final model. I don't think the data are suggesting that the direct effect actually came back. I'm assuming that there is some sort of error/inaccuracy because the direction of x-y changed to negative and it makes no logical sense that this M2 would've illuminated some hidden effect of x on y. I'm hoping there is some language i could use to explain this as a fluke. Any suggestions? Thanks again.