No announcement yet.
  • Filter
  • Time
  • Show
Clear All
new posts

  • Direct effects in mediation


    I have run a mediation model using SEM to test whether the relationship between the predictor 'psai_30mst' and the outcome 'ptsd_score' is mediated by the latent variables 'Self' and 'Mother' (while adjusted for mum_edu and ethnicity):

    sem (ptsd_score <- psai_30mst mum_edu ethnicity Mother Self) (Self Mother <- psai_30mst mum_edu ethnicity) (sexual_abuse_ever_combined_yn_re phsyical_abuse_ever_lifeevents_c emotional_abuse_ever_combined_yn phsyical_abuse_ever_parents_comb <-Mother) (ypb8002 ypb8003 ypb8006 ypb8007 ypb8000 ypb8001 ypb8004 ypb8005 ypb8052 ypb8053 ypb8056 ypb8057 ypb8050 ypb8051 ypb8054 ypb8055<-Self), latent(Self Mother) method(mlmv) stand
    The results of this are shown in the attached Image 1.

    I have then entered 'estat teffects' to get the direct, indirect and total effects. But the direct effects I get here are not the same as the results I got for these paths in image 1 (results shown in Image 2). So I am unsure which are the correct path coefficients.

    I have tried running 'estat teffects, stand' as I thought it might be due to standardization, but this gives the same results as 'estat teffects'

    I wondered if anyone knows why I might be getting different results here.


    Image 1:
    Click image for larger version

Name:	Image 1.PNG
Views:	1
Size:	23.7 KB
ID:	1478715

    Image 2:
    Click image for larger version

Name:	Image 2.1.PNG
Views:	1
Size:	7.6 KB
ID:	1478717

    Click image for larger version

Name:	Image 2.PNG
Views:	1
Size:	13.6 KB
ID:	1478716