What does simulating values from a known probability distribution mean? For instance, what does it mean to say, simulate values from $\text{Unif}(0,1)$? Does is mean that we have to construct a random variable $X$ such that $X\sim\text{Unif}(0,1)$, and we have to repeat the experiment many times to note down the values $X$ takes?


"Simulating values from a known probability distribution" means to generate realizations from a probability distribution. If you're talking about a uniform(0,1), this would mean generating a random number that has the same probabilistic properties as a random number that has a uniform(0,1) distribution, namely that $P(X \leq x) = x$, for $x \in (0,1)$

Simulating multiple realizations (or "repeating the experiment", as you put it) is a way to explore the properties of a probability distribution. For example, in R, if you wanted to look at variability in what a histogram and QQ plot of a sample of 50 realizations from a standard normal distribution looks like, you could paste the code

x <- rnorm(50)

several times.

Similarly, you could so a simulation study to explore variation in a sample statistic. E.g. if you wanted to explore what kinds of values you could see from the maximum of 20 uniform(0,1) random variables, you could type

max( runif(20) ) 

many times. You could similarly replicate these experiments a large number of times (say, 10000) to get a good estimate of the distribution of that sample statistic.

  • $\begingroup$ This is a very nicely written answer that should be helpful to new users like the OP. $\endgroup$ Jan 9 '17 at 6:30

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.