**k**, is small?

According the following link (Autocorrelation Function | Real Statistics Using Excel) the definition of autocorrelation is given by

with the following observation

Notice the autocovariance divides by the

__entire__count of the time series observations and the

__entire__mean of the time series observations. Whereas the regular definition of covariance is divided by the count of

__each__series and mean of

__each__series.

For large sets and small lags the consequences are minimized, but with few observations the effects are quite severe.