We briefly summarize the two most important objections of Yamaguchi and reply to them. Regarding the first point we are aware that Pearson cross- correlation between two autocorrelated series can lead to erroneous results. However, in the present case we preferred not to use a different approach because of the comparatively low quality and peculiar nature of the series analyzed. Such series are based on mostly subjective classifications and are very different from tree-ring width data. We performed a simulation calculating the Pearson cross-correlation of 10000 simulated couples of small (50 data) independent binary (0, 1 as possible outcomes) series, and count the number of significant correlations for different significant levels (the significance levels are calculated as in Cardaci et al. 1993). The results obtained show that the maximum difference of probability from what is expected is <1%. The results now shown confirm the ones obtained by Cardaci et al. (1993). Concerning the second point, we agree with the objection of Yamaguchi. For this reason we considered different time windows (not multiple ones of the others) and took only the stable results to minimize this problem. It is in fact very unlikely that a significant correlation merely due to chance shows up at the same lag in different windows.
|Titolo:||Replay to the comments by D.K. Yamaguchi on “Cross-correlation analysis of seismic and volcanic data at Mt. Etna volcano, Italy|
|Data di pubblicazione:||1995|
|Appare nelle tipologie:||1.1 Articolo in rivista|