Log doesn't unbound a 0-100 variable. You need something like logit, and be *very* wary of 0 or 100 values.
Something like
Ynew = .5/n + (1 - .5/n)*Y
Seems to work pre-logit.
-----Original Message-----
From: "Christian Weiss" <[email protected]>
To: [email protected]
Sent: 5/14/2009 3:13 AM
Subject: st: Bounded Dependent Variable in Random Intercept Model?
Dear statalisters,
I use a bounded variable ( 0-100%) as the dependent variable in
standard OLS regressions. To do so, I log-transformed the variable to
yield a unbounded variable.
Could you tell me if i have to use the log-transformed (unbounded)
variable for a random intercept model, or can I also use the initial
bounded variable?
best regards
Christian
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
