st: Balancing panel data uniquely identified by TWO variables

[Victor Gauto <vgauto@gmail.com>]

Dear  Statalist,

I have an unbalanced panel I would like to “fill” out.  I’ve tried
using the “tsset” and “tsfill, full” commands without getting the
results I wanted.  As an example, I have 2 companies selling 2
products in 3 years.  So if my data were complete I should have
2X2X3=12 observations to have a balanced panel.  I have fewer
observations, since some companies have not sold all the products in
all the years.  For example, I have 7 observations.  I would like to
fill this gap by including missing observations due to lack of sales
of certain products, or certain years.

The tsset syntax looks like this (I realize this may also be done with xtset):

"tsset panelvar timevar"

I think my panelvar is actually made up of two variables (company and
product; which are also unbalanced in my data), so I’m not sure how to
work around this.

I have tried the following:

sort year company_id product_id
gen panel_id=_n
tsset panel_id year
tsfill, full

I do not think this does it because "panel_id" above includes the year
dimension and goes from 1-7.  As a result, the code above OVERfills my
data.  Instead of getting the desired 12 observations, I get 7X3=21
observations, which is not what I want.

I want to go from this:

Year	Company_ID	Product_ID	Sales	Yearly Profits      n
1980	        1	                1	           250	      5000            1
1981	        1	                1	           225	      2000            2
1982	        1	                1	           120	      1200            3
1981	        1	                2	           125	      2000            4
1980	        2	                1	           250	      1600            5
1981	        2	                2	           450	      1700            6
1982	        2	                2	           300	      2200            7

TO THIS:

Year	Company_ID	Product_ID	Sales	Yearly Profits	n
1980	        1	               1	          250	              5000	1
1981	        1	               1	          225	              2000	2
1982	        1	               1	          120	              1200	3
1980	        1	               2	            .	              5000	4
1981	        1	               2	          125	              2000	5
1982	        1	               2	            .	              1200	6
1980	        2	               1	          250	              1600	7
1981	        2	               1	            .	              1700	8
1982	        2	               1	            .	              2200	9
1980	        2	               2	            .	              1600	10
1981	        2	               2	          450	              1700	11
1982	        2	               2	          300	              2200	12

Many thanks for any help,

Victor

*
*   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/