Thanks so much, Bill.
What the first part of your coding produced was exactly what I was
looking for. You are probably right that this not the best way to
store data. However, I have not been able to figure out how keeping
the data in the long form would make it easy for me to merge it with
other firm-level variables. Let's say I have data on two firm-level
variables "sales" and "profit" which look as follows:
Company Sales Profit
11A 100 10
11K 200 30
12Z 500 100
14T 400 150
21S 300 250
I can now easily merge this to the matrix, using the common variable
in both the data "Company". Now why do I want to combine these two
data sets? One reason. When I remove some firms from the data set
because, for example, they are too small (small volumes of sales), I
can easily make those changes in the matrix as well (because there is
one common company id for both the matrix and other variables). I'd
not in any case do matrix-based computations in Stata, but I'd use
some social network program to analyze the matrix and derive certain
network measures, which will then be brought back to Stata for
analysis, along with other firm-level variables. This is how I saw the
way to proceed, . But I've never used this type of data before, so I
think I am perhaps missing your point big time and therefore I'll
think further over your suggestion.
Thanks,
Joe.
On Thu, Jun 4, 2009 at 4:09 PM, William Gould, StataCorp LP
<[email protected]> wrote:
> I wonder whether Joe really wants that because I suspect that is not a
> useful way to store or analyze these data, but I may be wrong, so
> first, let's get the data organized in this way.
>
>
> Obtaining what Joe wants
> ------------------------
>
>
>
> +-------------------------------------------------+
> | company f_11A f_11K f_12Z f_14T f_21S |
> |-------------------------------------------------|
> 1. | 11A 0 0 2 1 0 |
> 2. | 11K 0 0 1 0 1 |
> 3. | 12Z 2 1 0 1 1 |
> 4. | 14T 1 0 1 0 0 |
> 5. | 21S 0 1 1 0 0 |
> +-------------------------------------------------+
>
>
> How I would organize these data
> -------------------------------
>
> I would organize these data in the long form, like this,
>
> +-----------------------------------+
> | c1 c2 company1 company2 n |
> |-----------------------------------|
> 1. | 1 2 11A 11K 0 |
> 2. | 1 3 11A 12Z 2 |
> 3. | 1 4 11A 14T 1 |
> 4. | 1 5 11A 21S 0 |
> 5. | 2 3 11K 12Z 1 |
> |-----------------------------------|
> 6. | 2 4 11K 14T 0 |
> 7. | 2 5 11K 21S 1 |
> 8. | 3 4 12Z 14T 1 |
> 9. | 3 5 12Z 21S 1 |
> 10. | 4 5 14T 21S 0 |
> +-----------------------------------+
>
*
* 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/
