Announcement

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

  • Calculating indirect effects with GSEM

    Hi,
    I am undertaking a path analysis of factors associated with nursing home placement.
    As my outcome is binary (nursing home placement) I am using the GSEM command in Stata 14.
    I would like to calculate the indirect effects of a mediator variable which is continuous (stress) on the binary outcome variable (nursing home placement).
    The Stata structural equation modelling reference manual (release 13) advises that the estat teffects command cannot be used with GSEM.
    It would seem that the indirect effect needs to be calculated by hand using nlcom. Have I understood this correctly or is there an updated command available that would calculate indirect effects with GSEM?
    Also, I need to obtain robust standard errors. Can this be done with nlcom?
    Any advice would be very much appreciated.
    Many thanks,
    Nora

  • #2
    Apologies I should of added to this post, can you use nlcom when one of the independent variables is binary? So path A is stress (continuous mediator) to nursing home placement (binary outcome). Path B is community care (binary independent) to stress (continuous mediator). Can I use nlcom to calculate the indirect effect by multiplying path A with path B, despite the mix of binary and continuous variables?

    Comment


    • #3
      noraadonnelly the problem here is with the scaling of the parameters. Path A would contain a parameter estimate in logits (presumably), while Path B would be in the metric of your mediator (stress). How would you then interpret the product of some quantity of logits and some quantity of stress? There's a related discussion in the context where the mediator is nominal and the independent variables are continuous in the Mplus User's Guide. There are ways to do it, but it isn't the easiest and I would imagine getting the correct standard errors could be a bit more challenging when you need some type of robust VCE.

      Comment


      • #4
        One way to do this in Stata is discussed here: http://maartenbuis.nl/publications/ldecomp.html
        ---------------------------------
        Maarten L. Buis
        University of Konstanz
        Department of history and sociology
        box 40
        78457 Konstanz
        Germany
        http://www.maartenbuis.nl
        ---------------------------------

        Comment

        Working...
        X