Speaking in terms of autocorrelation what is meant by "lags"?

Question

Speaking in terms of autocorrelation what is meant by "lags"?

Answer 1

The lag is the unit of time you are using in your autocorrelation. For example, suppose you were checking autocorrelation functions for an annual loon population. Your time unit is 1 year. If you have a high correlation for lag 3, then that means that the loon population is similar every third year; it suggests a 3-year cycle that can be further investigated using spectral analysis. If the lag 3 autocorrelation is strong, it should also show up in ACF values of lags 6, 9, etc., usually with some deterioration (erosion) as the lag values increase.

Keep in mind that in order to get meaningful ACF values, you first have to detrend your data. For population time series, you can use growth rates (lambda) or better yet, log-transformed growth rates.

Answer 2

the time where it appears that information has stopped being transfered, so that information can be built into real data.