Re: st: merging two datasets: companies with daily stock data

[Mitch Abdon <mitchaabdon@gmail.com>]

Hi,

It is possible to load a subset of a dataset. The syntax is:

 use [varlist] [if] [in] using filename [, clear nolabel]

See "help use" . In your case, I would think you will need the -if-
qualifier for the date conditions.

On Thu, Nov 11, 2010 at 1:59 AM, Nuno Soares <ndsoares@gmail.com> wrote:
> Hi everyone,
>
> I'm trying to implement a bit of code that I normally use in SAS, but I'm
> having some trouble to implement in Stata given the magnitude of data that
> it requires.
> Imagine you have two data files: one with a list of companies with a given
> date (say an event date) associated and the other with daily stock market
> data. The file with the companies can have the same company with multiple
> events. Problem: get the market data for a given period (say event date -
> 260 days).
>
> This problems seems to be easily solved by using a m:1 merge with the
> company ID as the merge variable. This would get all the daily data,
> irrespective of the event date, into the merge file and I could then delete
> those dates that are not needed.
>
> Now, imagine the company and event date file has 1000 observations all of
> which represent firms that have 10 years of daily market data, and I still
> just want event date - 260 days. This would mean that the merge process
> would lead to a file with 1000*10*260=2600000 observations, of which I only
> needed 10%. As both the files become increasingly bigger, the time needed to
> merge becomes longer and the memory requirements increase.
>
> In SAS I would only use a proc sql with the date restrictions, and it would
> get the data needed. In Stata, it seems that all the daily data file (with
> the restriction of the ID companies) is loaded into the memory and then we
> need to delete want we don't need. Is there a way of restricting the amount
> of daily data loaded into the memory in Stata using the  - merge - command,
> or a command that allows to do that? The 2600000 is not that bad, but I
> normally encounter 2 or three more times this number of observations...
>
> Best wishes,
>
> Nuno
>
> *
> *   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/
>



-- 
Best,
Mitch


Arnelyn Abdon
+639178034402
http://statadaily.wordpress.com

*
*   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/