Time series of observations generated by a stationary red noise process are characterized by long intervals when the observations remain above or below the overall mean value. These intervals can be easily misinterpreted as "climatic regimes" with different statistics. A "prewhitening" procedure that removes the red noise component from the time series prior to an application of a regime shift detection technique is discussed. The key elements of this procedure are subsampling and bias correction of the least squares estimate of the serial correlation. A new technique to obtain a bias-corrected estimate of the autoregressive parameter is proposed. It is shown that the Pacific Decadal Oscillation (PDO) appears to be more than just a manifestation of a red noise process.