Interpretation of mediation model (total effect not significant, but direct effect is)

[Anna]

Dear all, 

probably this is a stupid question - but I am confused how to interpret my mediation model testing two mediations using sem() of lavaan. 

I first tested the total effect (path c) with lm() and correlations, but it was not significant. However, I continued.

It appeared that path a and b are not significant, but now path c', the direct effect, was. 

This is the model, which converged without problems and showed to have a good fit. 

modelSad_TOTAL <- 'Correct ~ c1 * TOTAL

DurationFixation_Sad ~ a1 * TOTAL

Interference_Sad ~ a2 * TOTAL

Correct ~ b1 * DurationFixation_Sad

Correct ~ b2 * Interference_Sad

ab1 := a1*b1

ab2 := a2*b2

total1 := c1 + (a1*b1)

total2 := c1 + (a2*b2)'

and this is the output: 

Regressions:

                         Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all

  Correct ~                                                               

    TOTAL (c1)           -0.620    0.312   -1.984    0.047   -0.620   -0.177

  DurationFixation_Sad ~                                                      

    TOTAL (a1)            0.003    0.013    0.248    0.804    0.003    0.022

  Interference_Sad ~                                                          

    TOTAL (a2)            0.018    0.013    1.359    0.174    0.018    0.122

  Correct ~                                                               

    DurFix (b1)           -2.589    2.155   -1.201    0.230   -2.589   -0.106

    Interference (b2)      0.857    2.152    0.398    0.691    0.857    0.036

Variances:

                   Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all

   .Correct         561.793   71.930    7.810    0.000  561.793    0.957

   .DurFix            0.991    0.127    7.810    0.000    0.991    0.999

   .Interference      0.994    0.127    7.810    0.000    0.994    0.985

Defined Parameters:

                   Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all

    ab1              -0.008    0.034   -0.243    0.808   -0.008   -0.002

    ab2               0.015    0.040    0.382    0.703    0.015    0.004

    total1           -0.628    0.314   -2.000    0.046   -0.628   -0.179

    total2           -0.604    0.310   -1.949    0.051   -0.604   -0.173

Am I allowed to interpret this effect?

I would be happy, if one could help or give me any suggestions what to look up. 

Thanks in advance! 

Anna  

[Terrence Jorgensen]

I first tested the total effect (path c) with lm() and correlations, but it was not significant. However, I continued.

It appeared that path a and b are not significant, but now path c', the direct effect, was. 

This is not a lavaan issue.  SEMNET is a better place to ask general SEM questions like this

www2.gsu.edu/~mkteer/semnet.html

Terrence D. Jorgensen

Postdoctoral Researcher, Methods and Statistics

Research Institute for Child Development and Education, the University of Amsterdam

UvA web page: http://www.uva.nl/profile/t.d.jorgensen

[Anna]

You are right! I wanted to try my luck, though.. 

I will have a look into SEMNET, thanks a lot !! 

Best, 

Anna