Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | "Lachenbruch, Peter" <Peter.Lachenbruch@oregonstate.edu> |
To | "statalist@hsphsun2.harvard.edu" <statalist@hsphsun2.harvard.edu> |
Subject | st: RE: RE: GLLAMM tobit |
Date | Thu, 12 Jan 2012 13:23:56 -0800 |
I believe you will find informaiton on offsets in HIlbe and Hardin's book available thorugh the stata website. ________________________________________ From: owner-statalist@hsphsun2.harvard.edu [owner-statalist@hsphsun2.harvard.edu] On Behalf Of Bontempo, Daniel E [deb193@ku.edu] Sent: Thursday, January 12, 2012 1:21 PM To: 'statalist@hsphsun2.harvard.edu' Subject: st: RE: GLLAMM tobit I thought I would make a 2nd try to get help since I originally posted so close to the end of the Fall semester. I cannot find much information on scaled probit versus probit, or why scaled probit is used here since the cases with the SPROBIT link function will all be either 0 or 1. I am also not finding much about using an offset, outside of changing Poisson regression from count to rate. Any reference to a text discussing the use of offset for censored outcomes would be greatly appreciated. Thanks -----Original Message----- From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Bontempo, Daniel E Sent: Friday, December 16, 2011 11:22 AM To: 'statalist@hsphsun2.harvard.edu' Subject: st: GLLAMM tobit I am studying some code suggestions placed on StataList about tobit regression in GLLAMM. (http://www.stata.com/statalist/archive/2004-02/msg00789.html) If I could please ask a few questions about the offset. > * create a new dependent variable equal to 1 if right-censored > * and 0 if left-censored > gen y=cond(zteq>=100,1,cond(zteq<=0,0,zteq)) if zteq<. > * create offset variable equal to -100 if right-censored at > * 100, 0 otherwise > gen off = cond(zteq>=100,-100,0) 1) I understood offset values had to be in the units of the linear predictor. This offset of -100 seems to be in raw data units. 2) Also, I am not clear why it is negative 100, intuitively is seems like it should be +99. 3) Could this approach accommodate a non-zero floor? 4) Is this approach valid if the data are only left-censored (i.e., a floor effect)? There would be no instances of the new DV equal 1. (my data has a strong floor, but a good tail on the right.) 5) I have seen several references in GLLAMM manual, or publications, that state sprobit link is useful for floor/ceiling effects. But where can I read more about sprobit vs probit, and how it is used to address floor effects? Is there a key paper or text? Thanks -Daniel Bontempo * * 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/ * * 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/