Wang et al. reply.

REPLYING TO J. Carstensen, R. J. Telford & H. J. B. Birks 498, http://dx.doi.org/nature12272 (2013)Some issues have been raised with regard to our paper, by Carstensen et al.. In terms of our data processing, we were aware from the outset of the problems of unevenly spaced temporal data and sediment dating errors. We also wanted to duplicate, as far as possible, the methods published previously that had been used to identify early warning signals in palaeoenvironmental data (for example, ref. 3). Thus, we applied two standard smoothing functions (exponential and Gaussian kernel) to interpolated and non-interpolated (original) diatom data, expressed as three statistical indices (detrended correspondence analysis (DCA), Hill's diversity index N2 (HDI) and correspondence analysis), using different sliding-window sizes and the two-standard-deviation range of dates for each sample. [ABSTRACT FROM AUTHOR]