Re: st: "random" seed for uniform()

[David Kantor <dkantor@jhu.edu>]

Gary Longdon wrote, in response to Michael Oakes:


> [...]
> Then, I would argue against changing the seed at all.
> [...]
I would add one more thought.  If I am developing a program or do-file that 
uses random numbers, and I am tweaking the algorithm, then I want to see 
the resulting change that is due to each little change in the algorithm. I 
want to see only the effects of my programming changes -- not confounded by 
having a different sequence of random numbers. Thus, during development, I 
will set seed to the same value at the start of each run, to obtain the 
same "random" sequence each time. Once I am satisfied with the program, I 
may rerun it several times with different random seeds, such as those 
obtained from the time.

Similarly, if a procedure draws a random sample (or uses randomization 
combined with other methods to draw a selection) and I later want to 
retrace the process, but perhaps generating additional variables along the 
way, then I would want to use the same seed as was used before.

-- David

David Kantor
Institute for Policy Studies
Johns Hopkins University
dkantor@jhu.edu
410-516-5404

*
*   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/