Non-significance after mean-centering

Hi everyone,

I recently did a diary study in which my main variables revolved around incivility, emotions and sleeping problems.
I had one baseline survey which I asked for general levels of incivility in the past 3 months, demographics and personality questions. Then, the next week, I started with 2 daily questionnaires for 5 days- 1 in the morning and 1 in the afternoon.

I was testing a mediation. I put my within variables (incivility as X, emotional reactivity as M, sleeping problems as Y, and general incivility levels from the general survey as covariate).
When I test it, the mediation is only significant when my variables are not group-mean centered, and I really cannot understand why. I have tested using lme4 package but I also tried with the Macro PROCESS for R. I leave her the results with the last one.

The results when the variables are not mean centered is:
Running 1000 simulations...Indirect Path: "expcyberwi_mean" (X) ==> "emoreactivity_mean" (M) ==> "sleepingproblems_mean" (Y)
Effect S.E. z p [MCMC 95% CI]
Indirect (ab) 0.058 (0.012) 4.817 <.001 *** [ 0.036, 0.084]
Direct (c') -0.137 (0.030) -4.632 <.001 *** [-0.195, -0.078]
Total (c) -0.079 (0.028) -2.855 .004 ** [-0.132, -0.027]

When they are mean centered:
Running 1000 simulations...Indirect Path: "expcyberwi" (X) ==> "emoreactivity" (M) ==> "sleepingproblems" (Y)
Effect S.E. z p [MCMC 95% CI]
Indirect (ab) 0.006 (0.005) 1.180 .238 [-0.001, 0.018]
Direct (c') -0.071 (0.039) -1.841 .066 . [-0.146, 0.009]
Total (c) -0.065 (0.039) -1.687 .092 . [-0.142, 0.014]

I thought that perhaps my data has too many outliers. I tried simulating adding more responses and just test it, but everything remains the same.
From your experience is there any justification or is there anything I can do?


This topic was automatically closed 42 days after the last reply. New replies are no longer allowed.

If you have a query related to it or one of the replies, start a new topic and refer back with a link.