Announcement

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

  • Demonstrate Intermediate Variable is Significant

    Hi, first time poster here. But your threads have helped me a lot in the past!

    I am conducting a case-control study, and wand to demonstrate that risk factors a,b,c,d act through intermediate variable x to cause endpoint y. In my study, the a,b,c,d -> y relationship is well established but "x" is novel.

    For example, demonstrating risk factors hypertension, hyperlipidemia, smoking act through carotid atherosclerosis to cause stroke.

    I found a paper that achieves a similar goal, but the statistical methods were above my head. Free article here:
    http://pdfs.journals.lww.com/jaids/2...ZSejQtZjnaYQ==

    Thanks a lot!
    Last edited by Ben Stevenson; 19 Jul 2017, 12:01.

  • #2
    This is best done with structural equations modeling (SEM). If that is the statistical method that was above your head, then you're going to need to learn it before you can proceed. (The link you provided gave the abstract, but failed to show the full article. The abstract did not describe the statistical methods.)

    I highly recommend Alan Acock's book Discovering Structural Equations Modeling Using Stata, available from Stata Press. It gives a very readable introduction both to the theory and application of SEM, and also to the specific Stata commands that implement it. If you understand regression, this book will be at your level.

    Comment


    • #3
      Originally posted by Clyde Schechter View Post
      This is best done with structural equations modeling (SEM). If that is the statistical method that was above your head, then you're going to need to learn it before you can proceed. (The link you provided gave the abstract, but failed to show the full article. The abstract did not describe the statistical methods.)

      I highly recommend Alan Acock's book Discovering Structural Equations Modeling Using Stata, available from Stata Press. It gives a very readable introduction both to the theory and application of SEM, and also to the specific Stata commands that implement it. If you understand regression, this book will be at your level.
      Thanks! I will look that up.
      Fixed the link to go directly to the PDF.

      Comment


      • #4
        There are also several user-written routines that do mediation analysis (medeff, mediation, etc. - enter " findit mediation" from the command line). SEM is a more general method - the specific routines might be easier.

        Comment

        Working...
        X