Eunkyeong,
All you need to do is to drop the choice row for a particular individual if
that hospital is not in the choice set and code the hospital-specific dummy
variable as zero. So, re-writing your example and adding in a X variable to
help illustrate - here is what the data for your first two patients should
look like (assuming that there are only 5 hospitals - if there are more than
you need to create the appropriate dummy variables and code them to zero
when they don't match a row.
pat_id choice hosp X hosp_opt1 hosp_opt2 hops_opt3 hops_opt4 hops_opt5
1 1 1 5 1 0 0 0 0
1 0 2 4 0 1 0 0 0
1 0 3 2 0 0 1 0 0
2 1 1 5 1 0 0 0 0
2 0 4 8 0 0 0 1 0
2 0 5 10 0 0 0 0 1
and the correct command is 'clogit choice X hosp_opt2-hosp_opt5,
group(pat_id) - the conditional model can work out that some individuals
don't provide any information for distinguishing between particular choices.
BTW, unless your X's vary by both individuals and hospitals or you have
multiple observations per individual with different X's over time, you
cannot identify both hospital fixed effects and the coefficients on the X
variables
Cheers,
Steve
-----Original Message-----
From: [email protected]
[mailto:[email protected]]On Behalf Of Eunkyeong Lee
Sent: Wednesday, 12 September 2007 8:38 a.m.
To: [email protected]
Subject: st: clogit w/ different choice set
Hi-
Hi-
I want to estimate patient's hospital choice model, using clogit, but
my problem arise because each patient has different hospital choice
sets.
For example, suppose patient#1 chose hosp#1 although she could choose
hospital#2 and #3. Then, the data set should be arranged as follows:
Y(choice=1)=X's (hospital characteristics such as distance to hospitals)
and hospital-specific dummy variables.
pat_id choice some X's hosp_opt1 hosp_opt2 hops_opt3
1 1 1 0 0
1 0 0 1 0
1 0 0 0 1
The model would be
clogit choice X's hosp_opt2 hosp_opt3, group(pat_id)
However, now patient#2 chose hosp#3 when her choice set has hosp#1,
hosp#4, hosp#5.
pat_id choice some X's hosp_opt1 hosp_opt2 hops_opt3 hops_opt4 hops_opt5
2 1 0 . 1 0 0
2 0 1 . 0 0 0
2 0 0 . 0 1 0
2 0 0 . 0 0 1
Now I have to estimate
clogit choice(Y) X's hosp_opt1 hosp_opt4 hosp_opt5, group(pat_id)
So basically, for each patient, the number of hospital-specific dummies
is different as well as the identification of the dummies.
Could anyone help me out how to estimate clogit model in this case? I've
looked at the STATA manual book, googled it, and did everything, but I
could not still figure this out. Thanks.
Eunkyeong
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
