Re: st: duration analysis in gllamm

[Melaku Fekadu <melaku.fekadu@gmail.com>]

hi Steve,

thanks. you were very helpful.
i checked both. no special reason to prefer gllamm, i was just
referred to it. i checked the gllamm example on the data (cancer)
given on Jenkin's website,  it did not produce any result. the gllamm
example does not seem to include mass points option - if so how is it
coded?


melaku


On Mon, Jul 12, 2010 at 6:12 PM, Steve Samuels <sjsamuels@gmail.com> wrote:
> I second Maarten's suggestion. But, I ask: why -gllamm-?  I'd suggest
> you try -hshaz- or -pgmhaz8- by Stephen Jenkins, downloadable from
> SSC.  The setup is similar and the -help- for -hshaz- contains a
> -gllamm- example. See Chapter 9, especially Section 9.3, of Stephen's
> book "Survival Analysis" at
> http://www.iser.essex.ac.uk/files/teaching/stephenj/ec968/pdfs/ec968lnotesv6.pdf
> and his lesson 8 on setting up the analysis with Stata at
> http://www.iser.essex.ac.uk/survival-analysis
>
>
> Steve.
>
>
> On Mon, Jul 12, 2010 at 10:25 AM, Melaku Fekadu <melaku.fekadu@gmail.com> wrote:
>> Dear Statalisters,
>>
>> I want to estimate a duration model (time-to-first-employment) through
>> gllamm with unobserved heterogeneity. An individual may experience
>> transition in and out of states through years, as seen in below:  from
>> unemployment to employment, and from employment to unemployment. But,
>> I am interested only about the first transition from unemployment
>> (year 19XX=0) to employment (year 19XX=1).
>>
>> I wanted to ask two important questions about:
>> 1. the data structure for gllamm estimation
>> 2. the codes themselves in gllamm
>>
>> An example of the data is given below.
>>
>> Variables
>> Year 1996 =1 if employed in 1996, 0 other wise, and so on
>> X is some exogenous variable; it may be time-varying variable. In this
>> example it is not so.
>>
>> Data – Table 1, (my data currently structured as follows), each
>> individual has one row of observation with one entry for each year
>> For the first individual (row vector)
>> Id=1, year96=0, year97=0, year98=0, year99=0, year2000=1, x=12
>>
>> For the second individual (row vector)
>> Id=2, year96=1, year97=1, year98=1, year99=0, year2000=0, x=15
>>
>> For the third individual (row vector)
>> Id=3, year96=0, year97=0, year98=0, year99=0, year2000=0, x=10
>>
>> For the fourth individual (row vector)
>> Id=4, year96=0, year97=0, year98=0, year99=1, year2000=1, x=8
>>
>> For the fifth individual (row vector)
>> Id=5, year96=1, year97=1, year98=1, year99=1, year2000=1, x=17
>>
>>
>>
>> Restructured data (Table 2)
>> For the first individual – 5 rows of observation for each year
>> Id=1, Event=0, x=12
>> Id=1, Event=0, x=12
>> Id=1, Event=0, x=12
>> Id=1, Event=1, x=12
>> Id=1, Event=1, x=12
>>
>> For the second individual – 5 rows of observation for each year
>> Id=2, Event=1, x=15
>> Id=2, Event=1, x=15
>> Id=2, Event=1, x=15
>> Id=2, Event=0, x=15
>> Id=2, Event=0, x=15
>>
>> For the third individual – 5 rows of observation for each year
>> Id=3, Event=0, x=10
>> Id=3, Event=0, x=10
>> Id=3, Event=0, x=10
>> Id=3, Event=0, x=10
>> Id=3, Event=0, x=10
>>
>> For the fourth individual – 5 rows of observation for each year
>> Id=4, Event=0, x=8
>> Id=4, Event=0, x=8
>> Id=4, Event=0, x=8
>> Id=4, Event=1, x=8
>> Id=4, Event=0, x=8
>>
>> For the fifth individual – 5 rows of observation for each year
>> Id=5, Event=1, x=17
>> Id=5, Event=1, x=17
>> Id=5, Event=1, x=17
>> Id=5, Event=1, x=17
>> Id=5, Event=1, x=17
>>
>> Questions:
>>
>> 1.       If I want to use stata's gllamm, should I convert my data from that
>> of Table 1 to Table 2?
>>
>> 2.       Should I discard observations collected after the first transition
>> to employment has occurred? For example:  In case of individual number one,
>> observation 5 should be thrown? For individual number 2 (which is left
>> censored, because he is already observed working in the first period), which
>> observations should be thrown? Individual 3 is right-censored (has not yet
>> experienced employment at all), so should all of his observations remain in
>> the data? For individual 4, observation no 20 is collected after he has
>> already experienced employment in the previous period, so should it be
>> thrown? Individual 5 is left censored, so should his observations remain in
>> the data or be thrown?
>> 3.       If the data is to be restructured, for estimation through gllamm,
>> should the dependent variable be binary (one employed, 0 otherwise)? Or,
>> should it be a variable that indicates how many years has passed until the
>> individual became employed? For example: individual 1 is employed in the
>> fourth year from the beginning of the observation, so the variable takes the
>> value of 4; for individual 2, the variable takes value of 1 (since he is
>> already employed in the first observation)? And so on.
>> 4.        How should look like a code in gllamm unobserved heterogeneity
>> (parametric and non-parametric)? I will be grateful if you can indicate me
>> on how to code this in gllamm.
>> 5.  I would be very grateful if you have some codes of gllamm which would
>> give me some hints on how to code it.
>>
>>
>>
>> I really appreciate any help.
>>
>> Thanks a lot
>>
>> *
>> *   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/
>>
>
>
>
> --
> Steven Samuels
> sjsamuels@gmail.com
> 18 Cantine's Island
> Saugerties NY 12477
> USA
> Voice: 845-246-0774
> Fax:    206-202-4783
>
> *
> *   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/