1
$\begingroup$

I have the following time series data:

0.9803921569
0.9166666667
0.9090909091
0.8571428571
1.4915254237
0.4059620015

I want to check how correlated these data points are using autocorrelation. As far as I know, the lower the value of autocorrealtion, the more correlated the data. The issue that I can't understand is that everytime I change the order of the data, the autocorrelation value gets changed and that doesn't give me a clear indication whether they are correlated or not.

Can someone explain to me what I should do to fully understand how the autocorrelation works with these data? In other words, is there a pattern that the autocorrelation follows with such data to give me an indication whether they are correlated or not?

$\endgroup$
1
  • 2
    $\begingroup$ Autocorrelation has an explicit temporality; you are comparing a time series X(t) with a lagged version of itself X(t-1), X(t-2) etc. It is therefore appropriate that changing the order affects the autocorrelation since this destroys the temporality. I believe you can apply usual regression techniques, treating each lagged term as a new variable, though you don't have enough data for it to be meaningful and these should be indexed along some time/date $\endgroup$
    – Zac
    Sep 3, 2018 at 16:45

1 Answer 1

2
$\begingroup$

You only have 6 values? That in itself may make it harder to estimate auto-correlation among the values that you listed.

When are these values collected: every minute, every hour, every day, every month, every year, every decade, every century? Usually, the farther apart the values are from each other in time, the lesser the degree of auto-correlation. Conversely, the closer to each other the values, the higher the degree of auto-correlation.

Whether or not the values are collected at regular or irregular times will also determine how you assess the degree of auto-correlation.

Is the auto-correlation of interest in itself or just a nuisance? For example, it would be a nuisance if you were interested in estimating the trend in the collected values while accounting for their auto-correlation.

To sum up, we would need to know a lot more about your data to provide a pertinent answer.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.