I was thinking this, suppose you have a linear congruential generator that generates $n$ random numbers, we then make those integer numbers in doubles. So we have $n$ random numbers from $0$ to $1$.
Now suppose we need to normally distributed numbers, so we would input each number in a function that returns the quantile of a normal distribution associated with the number. Then we have our pseudo-randomly distributed numbers.
My question is: why do we use the inverse cumulative function of the normal distribution instead of the "ordinary" cumulative function?
Thank's.