Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, is already up and running.

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Re: st: binary mediation command

From   "Ariel Linden. DrPH" <>
To   <>
Subject   Re: Re: st: binary mediation command
Date   Mon, 4 Jul 2011 10:04:16 -0700

See the following website for a better understanding of how you use and
interpret binary_mediation.


Date: Sun, 03 Jul 2011 15:44:09 -0400
From: Pina Valle <>
Subject: Re: st: binary mediation command

Even when using percentile confidence intervals, is this output relevant
since I am trying to determine the proportion mediated by the indirect

Indirect effects with binary response variable evercoh
        indir_1 = .00601173        (adrel, binary)
total indirect  = .00601173
 direct effect  = .10223182
  total effect  = .10824356
       c_path   = .10680445
proportion of total effect mediated = .05553895 ratio of indirect to direct
effect  = .05880491 Binary models use logit regression


- ----- Original Message -----
From: Philip Ender <>
Date: Wednesday, June 15, 2011 4:21 pm
Subject: Re: st: binary mediation command

> Pina Valle wrote:
> >I am trying to test mediation with a dichotomous outcome, and I
> have looked around and found a command in >STATA called 
> binary_mediation. However, there isn't really any indication in the 
> notes I found on whether the >mediation is significant. Here is an 
> example of my output along with the commands that I have used:
> >...
> The documentation for -binary_mediation states:  This program does not 
> provide standard errors or statistical tests for coefficients.
> Bootstrap standard errors and confidence intervals are recommended for 
> this purpose.
> Using either the percentile or bias corrected confidence intervals, 
> whichever you prefer, intervals that contain zero are not significant 
> at the stated level while those that do not contain zero are 
> significant.  Traditional statistical tests are likely to lead to 
> biased p-values.  I think the confidence interval approach is safer.
> Phil
> --
> Phil Ender
> UCLA Statistical Consulting Group

*   For searches and help try:

© Copyright 1996–2015 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index