Trailing-window length.
Which fitted parameter to return: "mrr" = mean-reversion rate, "mean" = long-run mean, "relmean" = mean relative to current value, "std" = noise standard deviation.
Warmup behaviour.
https://screamer.readthedocs.io/en/latest/ for the Python reference and full details.
Rolling MLE fit of a mean-reverting Ornstein-Uhlenbeck process.