[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: RE: RE: fractional logit?

From   "Feiveson, Alan H. (JSC-SK311)" <>
To   <>
Subject   st: RE: RE: fractional logit?
Date   Fri, 1 Feb 2008 13:19:20 -0600

ALso consider using a beta distribution to model the fractions - 

ssc describe betafit

al Feiveson

-----Original Message-----
[] On Behalf Of Verkuilen,
Sent: Friday, February 01, 2008 12:47 PM
Subject: st: RE: fractional logit?

Viktor Slavtchev wrote:

>>I have individual data where people report the share of time they
in five distinct activities. People can do more than one activity every
day: first jogging, then go to work, then shopping, whatever. That is,
there are five observation per ID individual. Actually, the ID is the
(each person's) day. I wish to explore the individual characteristics
determining the share of time spent for each activity.
As the dependent variable is a share [0;1] I was thinking about
fractional logit. In Stata it would be (I guess):
glm y x, family(binomial) link(logit).
However, I am not clear whether fractional logit can be applied on the
original data, or the data have first to be transformed in some way.
Moreover, in doing -glm- live above I implicitly assume that each
observation is independent which is definitely not true.
Does anybody know some solution?<<

Yes, you are correct that there are notable dependencies among the
variables caused by the sum constraint. Take a look at the book by John
Aitchison, Compositional Data Analysis. The original edition is 1986 but
it was republished in 2003 with some additions. 

Feel free to drop me a line, I'm interested in this problem. 


*   For searches and help try:

*   For searches and help try:

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