Re: st: RE: RE: Encryption of data

["Rodrigo A. Alfaro" <raalfaroa@gmail.com>]

///
I sent this twice yesterday... it seems that it wasn't
distributed. Sorry for the "duplication".

Rodrigo wrote:
Very interesting discussion, I really like Maarten solution.
I was thinking in my lunch-time how to deal with the chance
of having ties in the uniform sequence (this could be important
for very large datasets). This is my small contribution.

(1) [Dealing with ties] Generate 2 or more uniform variables
and sort the main dataset with these. You can also add in the
sorting additional variables.

(2) [New key] Use Bill's idea of a long type of key. Save
the correspondence between id/key or (id time)/key in case
that you have a panel in a small dataset (such as code).

(3) [Other datasets] Hendri can impose the same keys than
before for other dataset (such as car).

Rodrigo.

/// Maarten's example (modified)
*------------ begin example -----------
sysuse auto, clear
set seed 12345    /// customized choice
gen double aux1 = uniform()
set seed 23456   /// customized choice
gen double aux2 = uniform()
sort aux1 aux2 price mpg    /// dealing with ties
gen long key = _n     /// Bill's idea.

preserve
sort key
drop make
save newauto    /// new dataset
restore

keep make key
sort key
save secret    /// codes

use newauto, clear
sort key
merge key using codes

use code, clear
sort make
merge make using car /// car is sorted by make
drop make
save newcar
*----------- end example ------------


----- Original Message ----- 
From: "Hendri Adriaens" <Hendri@uvt.nl>
To: <statalist@hsphsun2.harvard.edu>
Sent: Thursday, June 14, 2007 3:43 AM
Subject: RE: st: RE: RE: Encryption of data



> Hi William,
>
> Thanks for your information.
>
> > There is no additional security to be gained by doing that.
> > Ties do not
> > matter in this case.
> It might not matter for security, but for my application it does. The
> information from the master data set (that will be anonymised) will have 
> to
> be merged into a new dataset (to be anonymised with the mapping). If the
> mapping contains ties, -merge- wouldn't know which of the tied records to
> insert in the new dataset.
>
> But thanks, best,
> -Hendri.
>
>
> *
> *   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/ 
*
*   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/