0
$\begingroup$

Or is it maybe the same thing? I see that autocorrelation is when Yt is correlated with its lag Yt-1. But isn't that essentially what an AR process (say AR(1)) is? We are assuming that there IS correlation between its previous time period since we might see a directional trend from its initial data, right?

$\endgroup$
  • 2
    $\begingroup$ An autocorrelation is a function of a pair of times. An AR process is a stochastic process: thus, they aren't even remotely the same kinds of mathematical objects. On the face of it, then, your question makes no mathematical sense. Could you clarify what you're trying to ask? $\endgroup$ – whuber May 12 '19 at 19:53
2
$\begingroup$

Think of it this way:

If the index $t$ denotes time, then a stochastic process is simply a collection of random variables indexed by time.

There are stochastic processes (e.g., AR(1)) for which we can explicitly indicate how the value of $Y$ at time t (that is, $Y_t$) depends on values of Y at previous times. For example, an AR(1) process with mean zero stipulates that $Y_t = \phi*Y_{t-1}+ \epsilon_t$, where $\epsilon_t$ is a white noise process with zero mean and constant variance $\sigma^2$.

The autocorrelation function provides a measure of similarity between the values of $Y$ at times $t$ and $s$ by computing the correlation between $Y_t$ and $Y_s$, namely $Cor(Y_t, Y_s)$. (For a weekly stationary stochastic process, this correlation only depends on how apart in time $t$ and $s$ are from each other, it does not depend on the actual values of $t$ and $s$.)

For the above AR(1) process, the correlation between $Y_t$ and $Y_{t-1}$ is given by:

$Cor(Y_t, Y_{t-1}) = Cor( \phi*Y_{t-1}+ \epsilon_t, Y_{t-1}) = \phi$

$Cor(Y_t, Y_{t-2}) = \phi^2$

and, more generally,

$Cor(Y_t, Y_{t-k}) = \phi^k$, where $k > 2$ is an integer. (See https://www.mathstat.dal.ca/~stat5390/Section_3_ACF.pdf.)

$\endgroup$
  • $\begingroup$ Your opening characterization of stochastic processes seems quite at odds with standard conceptions of them. Did you perhaps mean to describe AR models? $\endgroup$ – whuber May 12 '19 at 21:17
  • 1
    $\begingroup$ Thanks, @whuber! I made some edits to the original answer to reflect your concerns - let me know if they look fine to you. I want to keep things simple. $\endgroup$ – Isabella Ghement May 12 '19 at 21:24

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.