How should we select sample points for estimating autocovariance with lag h in an AR(1) model?

A good way to select sample points for estimating autocovariance with lag h in an AR(1) model is to use the stationarity property of the series. To do this, select sample points such that their corresponding lagged values are equally separated (i.e., equally spaced with the same lag). That way, we can ensure that the autocovariance estimates are not influenced by an underlying trend or other non-stationary features in the series. This also allows us to estimate the autocovariance at different lags while taking into account that the exact sample point used is not necessarily the one at that precise lag.