Message/Author |
|
Ana Soper posted on Sunday, October 21, 2007 - 6:43 pm
|
|
|
How does one use Mplus to assess the significance of an indirect path in path analysis through the multivariate delta method, specifically? As a follow-up, can one both bootstrap, use type=missing h1, and use the delta method all in one analysis? Thank you very much for your help. |
|
Ana Soper posted on Sunday, October 21, 2007 - 9:46 pm
|
|
|
Maybe I'll try to answer my own post...is what is listed under "Specific Indirect" here the answer to significance of an indirect path using the Delta Method? Specifically, is the "estimate" provided under "Specific Indirect" akin to the t-test indicating significance for the Delta Method? TOTAL, TOTAL INDIRECT, SPECIFIC INDIRECT, AND DIRECT EFFECTS Lower .5% Lower 2.5% Estimates Upper 2.5% Upper .5% Effects from C1NEDIS to ALCDIS Sum of indirect -0.029 -0.018 0.002 0.036 0.058 Specific indirect ALCDIS C2THDIS C1NEDIS -0.029 -0.018 0.002 0.036 0.058 |
|
|
The results you show come from the MODEL INDIRECT command. The default standard errors are the Delta method. You can also ask for bootstrap standard errors. This can all be done under missing data estimation. The number in the third column is a z score. |
|
Ana Soper posted on Wednesday, October 24, 2007 - 4:03 pm
|
|
|
Thank you for your helpful reply. Just to clarify - if I want to bootstrap yet also obtain the standard errors for the Delta Method (default) rather than the bootstrap standard errors to evaluate significance, would I just type: OUTPUT: residual sampstat TECH1; meaning that I only want to see the standard errors for the Delta method and not bootstrap standard errors - rather than OUTPUT: residual sampstat TECH1 standardized cinterval (bcbootstrap); ? Thank you very much! |
|
|
If you don't specify the BOOTSTRAP option in the ANALYSIS command, you will not get bootstrap standard errors. |
|
Back to top |