I have a question about the basic definition of discrete normal distribution.
Let's assume I have a machine that draws a number ranging from 1 to 3 from a uniform discrete distribution (the probability to get 1, 2 and 3 is equal and equals to 1/3).
If I want to use that machine to create another machine that yields from the set {1,2} in a uniform manner, can I use the first machine in the following way:
draw a number -> if the number is 3, draw again. if not, return the number.
on one hand, it does seem like the new machine draws uniformly from {1,2}, as the two numbers have an equal probability of being drawn.
on the other hand, it's not a classic uniform distribution, as it might take a long time to draw the number (if the numbers being drawn is 3 in many trials).
So, can I say my new machine draws uniformly from the set {1,2}? thank you!
Yes you can. There is nothing in the definition of the uniform distribution that says anything about how it might be implemented or arrived at.