re: Re: re: st: mediation analysis with proportional data

["Ariel Linden" <ariel.linden@gmail.com>]

Per the -khb- help file: " model-type can be any of regress, logit, ologit,
probit, oprobit, cloglog, {help slogit}, scobit, rologit, clogit, xtlogit,
xtprobit and mlogit. Other models might also produce output but for the time
being this output is considered to be "experimental".

Thus you could (a) try a different model as assume the results to be
"experimental" or you could (b) use regress on your data and be cautious
with values the boundaries of 0 and 1.0, or you could (c) transform the
variables as suggested in the following faq:

http://www.stata.com/support/faqs/statistics/logit-transformation/

If none of those options work, you may want to email Kristian Karlson (the
"K" in -khb-) and ask his opinion on using any of these approaches or
others, since he's spent a lot of time investigating mediation models on
various data types.

I hope this helps

Ariel
Date: Fri, 14 Feb 2014 11:07:40 +0000
From: "Sandra Virgo" <Sandra.Virgo@lshtm.ac.uk>
Subject: Re: re: st: mediation analysis with proportional data

Hello Ariel

Thanks for your help, in particular the information about the khb method. 

The variables in question are not actually binary - they are proportional
values, so they range between 0 and 1 with lots of different values in
between. 

Using medeff did not work, presumably because its logit command only works
with 0 and 1 values. 

Does the khb programme work with generalised linear models like the
fractional logit I previously used on these data i.e. is it suitable for
continuous values bounded by 0 and 1? 

Thanks
Sandra Virgo
London School of Hygiene & Tropical Medicine



*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/