Hi Mr. Greenberg,
Thanks for your comments, especially pointing out the potential problem
of age and year being perfectly confounded. I'll keep it in mind.
Thanks!
Mandy
David Greenberg wrote:
>Yes, you can do that. You should understand, though, that if everyone
is 18 in the first year of your study, that age and year will be
perfectly confounded. Consequently you will not be able to distinguish
age effects from year effects. Depending on the subject of your study,
that may or may not be a problem. David Greenberg, Sociology
Department, New York University
>
>----- Original Message -----
>From: Man Jia <[email protected]>
>Date: Sunday, September 21, 2008 12:10 pm
>Subject: st: panel data structure using fixed effect model
>To: [email protected]
>
>
>> Hi all,
>>
>> I was wondering if I could get some help about a general econometric
>>
>> question of structure of panel data using fixed effect model. Sorry
>> that this question is not particular to any stata commands. I'm
>> thinking some of our experts on statistics or econometrics could give
>>
>> me valuable comments.
>>
>> I'm working with a panel data set of a group of individuals who have
>> different ages in the beginning. My question is: can I do the fixed
>> effect according to individual and age,like:
>>
>> Y[i,a]=X'[i,a]B + r [i] + t[a] +u [i,a] (1)
>> where
>> i means individual
>> a means age
>> t means year dummy at different ages
>> r means individual fixed effect
>>
>> Accordingly, the data set will be like this:
>> ---------------------------------------------------------------
>> ID age year var1 at age18-20 var2 at age 18-20
>> 1 18 1980
>> 1 19 1981
>> 1 20 1982
>>
>> 2 18 1978
>> 2 19 1979
>> 2 20 1980
>>
>> 3 18 1982
>> 3 19 1983
>> 3 20 1984
>> ----------------------------------------------------------------
>> The reason that I have the idea of trying (1) is that my interest is
>> age effect and I want to restrict the age in the range [18,20].
>>
>> The above data structure is kind of weird since the usual way is like
>>
>> this:
>> Y[i,t]=X'[i,t]B + a[i,t] + r [i] + t +u [i,t] (2)
>> Where i,t,a,r means the same as above( a means age)
>>
>> The usual data set correspondent to (2) is:
>> ---------------------------------------------------------------
>> ID year age var1 in year80-82 var2 in year 80-82
>> 1 1980 18
>> 1 1981 19
>> 1 1982 20
>>
>> 2 1980 20
>> 2 1981 21
>> 2 1982 22
>>
>> 3 1980 16
>> 3 1981 17
>> 3 1982 18
>> ----------------------------------------------------------------
>> or something like this:
>> ---------------------------------------------------------------
>> ID year age var1 in year80-82 var2 in year 80-82
>> 1 1978 .
>> 1 1979 .
>> 1 1980 18
>> 1 1981 19
>> 1 1982 20
>> 1 1983 .
>> 1 1984 .
>>
>> 2 1978 18
>> 2 1979 19
>> 2 1980 20
>> 2 1981 .
>> 2 1982 .
>> 2 1983 .
>> 2 1984 .
>>
>> 3 1978 .
>> 3 1979 .
>> 3 1980 .
>> 3 1981 .
>> 3 1982 18
>> 3 1983 19
>> 3 1984 20
>> ----------------------------------------------------------------
>>
>> Please let me know if there's anything unclear of my expression of the
>>
>> question. So, is there any problem of using fixed effect like (1)?
>>
>> Any comments will be very helpful. I appreciate it. Many thanks!
>>
>> Mandy
>>
>> *
>> * 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/
