Talk:Convergent cross mapping

Update for CCM
Dear Wikiverse,

After review of the CCM page, based on recent developments, and, to clarify and improve the citations, I am proposoing to update the page.

For instance, citation of an opinion article from a non-peer reviewed magazine does not meet Wikipedia standards for peer-reviewed scientific literature as the basis of objective article references. The citation of Cenys is not quite appropriate as this paper does not relate to convergent cross mapping, rather to estimation of interaction coefficients.

I also propose to rewrite the Theory section. For instance, the use of "shadow" manifold is ambiguous. In fact, there is no requirement for a "shadow" manifold, the algorithm works equally well on an observed multivariate state-space.

The algorithm section can be improved. "Creating weights" is an internal aspect of the simplex projection for which there is no parametric control, thereby it is not really an actual "step" the user undertakes. A crucial and often ignored aspect of CCM is the random sampling (which is the aspect that qualfies CCM as "statstical test") that averages the result of state-space projected cross mappings. In fact, the last step listed in the algorithm is not quite right, as this averaging is done over the Pearson correlations of each ensemble. Also, significance testing is an important aspect for practical use of the metric.

There have also been significant extensions to CCM in the last few years. These should be cited.

Thank you!

J Park (talk) 03:47, 14 September 2022 (UTC)