1
$\begingroup$

This is for a simple psychology report.

I was wondering if I could hypothesise an insignificant mediation.

Eg: "...based on this research it is predicted that group identity is not a significant mediator (or does not sognificantly control) narcissism's relationship with aggression."

Then I would run a mediation analysis using indirect syntax for SPSS and report that narcissism has a significant relationship with aggression, which is not significantly influenced by (does not significantly overlap with) group identity.

Or should i just include it in a regression and that would be clearer? But then my main question remains, can I predict in my hypothesis that a variable will have an insignificant correlation value? Or a weak relationship?

I basically want to say that i predict that narcissism will interact with rumination on aggression (using modprobe syntax and correlational data) . But i would also like to add in my initial prediction (based on research) that narcissism's relationship with aggression has nothing to do with group identity.

Thank you all for any input. As I said this is a basic psychology report, not worrying too much about null hypothesis and too much other hardcore statistical stuff. I just need a basically worded prediction (which i think i have) and a basic way to test it.

So the basic question is: can I speciffically predict insignificant correlations, mediations or even interactions? And would confirming this prediction support the notion that the two are unrelated in their relationship with aggression? (Or even that group identity has no relationship with aggression, not even via narcissism?)

Thank you for reading.

$\endgroup$
7
  • $\begingroup$ What kind of data do you have? What you could do is first state your hypothesis, then run a regression with aggression as your dependent variable and narcissism as your regressor (possibly among others). Then you could see from that regression whether the variable is significant or not: i.e. whether narcissism indeed has the effect on aggression that you predict. $\endgroup$
    – rbm
    Commented Jul 1, 2014 at 17:36
  • $\begingroup$ Thank you, but the yes ive done the narcissism on aggression, and it does have the predicted effect. Now i would like to modify my prediction to say this effect has nothing to do with group identity and i was wondring if a mediation which shows group identity as insignificant would support my prediction (or confirm it). Thank you very much for the input. $\endgroup$
    – red
    Commented Jul 1, 2014 at 18:49
  • $\begingroup$ Add group identity in the regression and see if that is significant. Also perform another regression of narcissism on group identity to see if there is an indirect effect from group identity on narcissism which in turn affects aggression. $\endgroup$
    – rbm
    Commented Jul 1, 2014 at 18:53
  • $\begingroup$ Also its correlational data, collected on questionnaires via scales of 4, 6 and 7 points. Thank you. $\endgroup$
    – red
    Commented Jul 1, 2014 at 18:54
  • $\begingroup$ Sorry for being slow. But would the last regression be the same as running the indirect mediation analysis syntax and reading the indirect path. And if it is an insignificant effect in the last regression than that would confirm the prediction that narcissism's relationship with aggression has nothing to do with group identity? Thank you very much. I think i just about get it, as well as i need to anyway lol. $\endgroup$
    – red
    Commented Jul 1, 2014 at 18:59

1 Answer 1

1
$\begingroup$

Yes.

If you wish to provide evidence of the absence of an effect, you should perform an equivalence test (e.g. using TOST) on the estimated parameter or combination of parameters representing your mediation.

$\endgroup$
1
  • $\begingroup$ Thank you, i will look into it, the link you provided is a bit complex for me (sorry total noob, can only really do stats through computer programs and have no idea about the equations behind it all, i know you guys who are passionate about stats probably hate people like me). But i will try to find a simple macro or simple instructions on how to do that (equivelence test) in spss. Thank you $\endgroup$
    – red
    Commented Jul 1, 2014 at 18:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.