Hi Statalist,
Have a question on the design matrix of a fixed
effects model.
I have cigarette sales and price data for a panel of
ten provinces. Some provinces have data for 19 years
(1981-1999) while other have data for 14 years
(1981-1989, 1995-1999. missing at random).
I regress sales on price, other provincial level
covariates, provincial and year dummy variables, and
provincial specific time trends. My question is on the
last item.
To create provincial specific time trends variables, I
create a time trend variable, and interact it with
provincial dummy variables. For a province with full
data, value of the trend variable are 1, 2, 3,... 19
corresponding to years 1981 to 1999.
For a province with only 14 data points, the trend
value for 1981 to 1989 are the same as above. However,
I do not know the proper way for values of 1995 and
thereafter. I can think of 2 possible ways:
Year Way I Way II
1981 1 1
1982 to 1988 skip here skip
1989 9 9
(1990-1994) missing data missing data
1995 15 10
1996 16 11
1997 17 12
1998 18 13
1999 19 14
Which way of coding is correct? Does the coding depend
on the research question in hand? I have searched on
some fixed effects model with unbalanced panel and
unequal spacing, and found nothing said on this.
Thanks very much for all of your help.
Rita Luk
Research Officer
Ontario Tobacco Research Unit
University of Toronto
Toronto, Ontario, Canada
(416) 535-8501 x4727
____________________________________________________________________________________
Be a better friend, newshound, and
know-it-all with Yahoo! Mobile. Try it now. http://mobile.yahoo.com/;_ylt=Ahu06i62sR8HDtDypao8Wcj9tAcJ
*
* 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/
