I have a problem with usage of expectation in Central limit theorem.As example, look at this problem:

A certain component is critical to operation of an electrical system and must be replaced immediately upon failure. If the mean lifetime of this type of component is $100$ hours and standard deviation $30$ hours, how many of these components must be in stock so that probability that system is in continual operation for the next $2000$ hours is at least $0.95$?

As we know: $E[\frac{(X_1 + X_2 + ... X_n)}{n}] = u$

BUT the solution to this problem is: $P( Z < \frac{2000-100n}{30\sqrt{n}}) = 0.05$, where $n$ is the number of components.

So the question: Why do we multiply our mean ($100$ hours) by $n$? And in which cases do we not multiply the mean by $n$?



Note that

$$ \sqrt{n} \frac {\frac {1} {n} \sum_{i=1}^n X_i - \mu} {\sigma} = \frac {\sum_{i=1}^n X_i - n\mu} {\sqrt{n}\sigma}$$

So both form are equivalent.

LHS has a better interpretation as the terms inside the fraction are finite, and the fraction converge to $0$ by LLN. But CLT tell us that by scaling with $\sqrt{n}$, the fraction does not goes to infinity; instead it has a normal distribution which is finite almost surely.


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.