[salma ktat]

I wrote some code a while back to do -infile- on a monster text file containing strings, that not only made it possible, it ran in a fraction of the time they eventually accomplished by bumping the server up to 64 GB. If you have a list of codes, it might look something like:

foreach code in ABCDA ABCDB ABCDC ABCDD {
use myfile if ID== `code'
sort date
[do stuff to get your annualized returns, I dunno if you compound]
keep [annual returns? important variables? regardless cut things down, cases and variables]
save temp_`code', replace
}

then run another loop to append the important parts.

I do think that Sergiy is correct, but if you can't re-install Stata, a work-around of the type I describe *might* get the job done.

Hi Ben,
Thanks for your suggestion. But as I'm a beginner with stata, I can't see what you mean by "If you have a list of codes". what codes am I supposed to write? Sorry that may seem a basic question but I really appreaciate any help with that. Thanks.

[Clyde Schechter]

So, find your original installation discs. You should be able to reinstall Stata, selecting the 64 bit version this time, using the same licensing information you currently have. (I know that is true for Stata versions 12 and 13; I believe that back in Stata 11 the license was not restricted to any particular operating system.)

[ben earnhart]

By "list of codes" I meant what in your case would probably be stock ticker symbols, which in your data, is probably the "ID" variable.

You might be able to extract that with "use ID using myfile" since even with a monster file, one variable doesn't take up much memory,

That said, the work-around is a desperation move, and best bet is to re-install with a 64-bit version of Stata, Which still doesn't solve things entirely (6.5 GB file with 4 GB RAM), but gives additional flexibility, like, for example, only reading in the variables you need.

[Sergiy Radyakin]

Salma, the best you can expect from Stata in such a setup is 1.2-1.3GB as per email of Bill Gould:
http://www.stata.com/statalist/archi.../msg01125.html

So, find your original installation discs.

And the corresponding original license.

This should probably close this thread.

Sergiy

[Attaullah Shah]

I think ben earnhert solution is wonderful. Salma you can try it. What he means is that you accomplish the job in parts, not whole at once. For example, you can calculate stock returns and annual returns say for 10000 firms, save those results in temporary file 1, then repeat the process for remaining firms. Once all the results are ready, combine the results files and drop non-required variables.
In your second post, your example data set has many redundant variables such as date coded as string as well as date format. you may want to keep only of them.

[Attaullah Shah]

Also, Salma you are using a cracked version of Stata, which is not legal

[Carlo Lazzaro]

Salma:
as an aside, a short cautionary tale. Some years ago a post across the list warned about not putting Stata licence serial number in public domain in order to avoid possible abuse by persons obviously outside the list.

Kind regards,
Carlo

[Sergiy Radyakin]

Carlo, could you please provide the citation for the post you mentioned? As far as I know it is safe to show the information displayed in -about-. Here is Alan Riley's serial number:
http://www.stata.com/manuals13/u5.pdf
(page 2), similarly the manual for 11th version contains a serial of Chinh Nguyen;Bill Gould has revealed his serial in Stata Journal, and serials of Pete Huckelba and Brian Poi (some nice small numbers) have also appeared in internet. They are also flashed during the Stata user group meetings, so I believe they are genuine serials, not editor-replaced with random numbers.

And every log file of the batch-mode session would contain the same number, so what is the rationale for hiding it?

As one of my projects depends on this, I really need to know whether there is any reason for hiding the serial number (as shown in the output of the about command)? and what possible abuse are you talking about?

Code and authorization that are entered during the installation and never displayed again by any Stata command are obviously private information, which should only be revealed to StataCorp's personnel. Don't confuse them with the serial number reported by the about command.

Thank you, Sergiy Radyakin

[Ingrid Lambert]

Dear Statalist experts,
I also had the same issue with the 32bit Stata14 running on 64bit Windows 8.1 (8GB RAM)
It wasn't possible to allocate more than 1GB of memory, virtual one did not help either.

Iteration 9:   log pseudolikelihood = -425.83053  
op. sys. refuses to provide memory

After an upgrade to a 64bit Stata this issue disappeared.