a1 = .006 (.006) a2 = .059 (.051) g1 = .041 (.018) g2 = .061 (.055) d1 = 3 years (37 years) d2 = 73 years (84 years) Ci = -.041 degs Residual power spectral density (peak is at 57.9 years)
Reverse TSI Bootstrap
Ran the sim with the reversed time series below
Using the parameters from the prior optimization correlation is negative as expected.
First step of the optimization (match the acfs) works because it is insensitive to scaling, including a scaling of -1
The fit though is not too is not too hot 🙂
Skip the first step and give the optimizer control of all parameters, best fit (not guaranteed global minimum)
‘course now the acf is horked
Leif’s magic series
The results at the top of this page were manually optimized. Running the optimizer does much better.
Run the sim to 1976