The ‘set seed’ command tells Stata where to begin drawing
random numbers from. In order to understand this you must keep in mind that
Stata cannot generate truly random numbers but can only draw numbers that look
random. (This is the limitation of binary computers. If Stata were running on a quantum computer
then truly random numbers could be possible.)
There is a benefit in this however. Numbers that look random but are not but are sufficiently close to being random to give us the ability to make inference about how a random number would work. The true benefit of not dealing with random number is that these numbers can be replicated.
This is extremely convenient in that it allows two researchers
in two different parts of the world to run the same simulation and get the
exact same results. In general the
exactness of the results is trivial.
However, it is extremely useful to know that you can change parameters
this way or that way, but that the underlying 'random' draw remains the same. Likewise, if a researcher in a different part
of the world were to run the same simulation with different parameters, it is
likely that the differences between the two simulations is caused by the change in
parameters, not the change in the random draws.
No comments:
Post a Comment