Figure 3 is the one where they tried different time intervals, ahead or in the past, and came up with 15 days ahead. The fine-tuning of the time parameter to produce 6σ result reminded me of https://xkcd.com/882/, however in the paper they say they’ve applied same parameter to a different dataset from earlier years and can still detect the correlation.
Did I read that right that it gives a 15 day lead time??
Figure 3 is the one where they tried different time intervals, ahead or in the past, and came up with 15 days ahead. The fine-tuning of the time parameter to produce 6σ result reminded me of https://xkcd.com/882/, however in the paper they say they’ve applied same parameter to a different dataset from earlier years and can still detect the correlation.