Adding capabilities

[CommonClimate]

There is interest in adding functionalities to Ammonyte to detect tipping points (broadly defined). There are several methods to choose from:

• Augmented Kolmogorov-Smyrnov test . Matlab code available here -- needs translation to Python (use an LLM!)
• LERM as used by our very own @alexkjames , e.g. on synthetic and real data, like ice core and speleothem d18O. The LERM code is available in Python via the Ammonyte package.
• Bayesian change point detection, e.g. that implemented in Ruggieri & Antonellis, 2016.
• Frequentist change point detection (scientifically more shaky, but there is a very nice package for it, called ruptures).

And more that I either missed in my review of the literature, or still in the works, e.g. this paper, still in review.

This might be a good project for @niatimary to get her feet wet with Ammonyte, and this will be an important component of PaleoPAL, so it is a good foundation to build now.

Let's start with KS test and go from there.

[CommonClimate]

Another potential method to try: Erhardt et al (2019), which is Bayesian, but involves Monte-Carlo Markov Chain simulations, and will therefore be slower than closed-form solutions like Ruggieri & Antonellis (2016). However, unlike the latter, it was designed for paleo and has been demonstrated on NGRIP data. Good news: it is already coded in Python