I have a question about the CLT. Suppose we have the independently and identically distributed random variables $X_i$ with mean $\mu$ and variance $\sigma^2$. Then, by the Central Limit Theorem

$$ \sqrt{n}\left(\frac{1}{n}\sum_{i=1}^n X_i-\mu\right) \rightarrow N\left(0, \sigma^2\right)\Longleftrightarrow \sum_{i=1}^n X_i \rightarrow N(n\mu, n\sigma^2), $$ as $n\rightarrow \infty$. But what if we have $\sum_{i=1}^nc_iX_i$, where $c_i$ are constants? Is it simply to say that

$$ \sum_{i=1}^nc_iX_i \rightarrow N\left(E\left[\sum_{i=1}^nc_iX_i\right], \;V\left[\sum_{i=1}^nc_iX_i\right]\right), $$

that is,

$$ \sum_{i=1}^nc_iX_i \rightarrow N\left(\mu\sum_{i=1}^nc_i,\; \sigma^2\sum_{i=1}^nc_i^2\right), $$

or am I missing something here?

  • $\begingroup$ What is the meaning of a convergence in distribution to a normal distribution with variable mean and variance? The question applies to the convergence to $N(n\mu,n\sigma^2)$ at the beginning of the question as well as to the convergence you are trying to establish later on. $\endgroup$ – Did Apr 14 '13 at 11:35
  • $\begingroup$ I'm not sure I understand your question. A sum (and a sample mean) converges to $n$ times variable mean and variance, no? Or do you mean that it has to be expressed in terms of convergence to a standard normal? $\endgroup$ – hejseb Apr 14 '13 at 12:12
  • $\begingroup$ In probability as in other branches of mathematics, the statement that $x_n\to y_n$ has no meaning. The convergence of a sequence $(x_n)$ is expressed as $x_n\to y$, for some $y$ which does not depend on $n$. The two statements my previous comment addresses both fail this requirement. $\endgroup$ – Did Apr 14 '13 at 12:16
  • $\begingroup$ Ah, yes, of course. That makes sense and is something I didn't think about. Thank you for your help, and excuse my mathematical recklessness! $\endgroup$ – hejseb Apr 14 '13 at 12:24

You might be after central limit theorems à la Lindeberg or Lyapunov, see here. These yield the convergence you are interested in, but stated more correctly, for example, some conditions ensuring that the sequence of general term $$ \frac{\sum\limits_{i=1}^nc_iX_i-\mu\sum\limits_{i=1}^nc_i}{\sigma\cdot\sqrt{\sum\limits_{i=1}^nc_i^2}} $$ converges in distribution to a standard normal distribution.


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.