How to estimate a probability based on the difference of two random variables?

231 Views Asked by At

There are two random variables:

$X ={}$ the sum of $1,000,000$ real numbers
$Y ={}$ the same sum as $X$ with each number rounded to the nearest integer before summing

If the fractions rounded off are independent and each one is uniformly distributed over $\boldsymbol{(-0.5, 0.5)}$, use the CLT to estimate the probability that $|X-Y|>700$.

My thought process so far is that $X \thicksim \operatorname{Uni}(-0.5, 0.5)$ and $Y \thicksim \operatorname{Uni}(-1, 1)$, but since $Y$ depends on $X$, what would my actual definition of $Y$ look like? How do I figure out the difference between a uniform and discrete variable?

I think I understand what I am supposed to do once I figure out the correct normal variable to use the CLT on, but I'm just not quite sure how to get there. Any help is appreciated.

1

There are 1 best solutions below

0
On

From my understanding of the question, I would set it up like this: we have $\{ X_i \}_{i=1}^{1000000}$ fixed deterministic numbers, which we sum $X = \sum_{i=1}^{1000000} X_i$, and then we wish to compare that with the sum $Y = \sum_{i=1}^{1000000} Y_i$ is, where $Y_i = \left[ X_i \right]$. We make the modeling assumption that the errors are a random variable $Z_i = Y_i - X_i \sim U(-0.5,0.5)$. Now we want to calculate $$P(|Y - X| > 700) = P(|\sum_{i=1}^{1000000} Y_i - X_i| > 700)$$ $$= P(|Z| > 700)$$ where $Z$ is a sum of one million uniform random variables and should obey the CLT. From here, we use our knowledge that $Z$ behaves like a normal distribution with mean 0 and variance $\frac{1000000}{3}$.

Hope that helps.