orbeckst
commented
5 years ago As discussed, try to make the parallel algorithm from one of the following work:
-
Philippe P. Pébay: "Formulas for Robust, One-Pass Parallel Computation of Covariances and Arbitrary-Order Statistical Moments." Technical Report SAND2008-6212, Sandia National Laboratories, September 2008. PDF Equations 1.3 and 1.4 describe how to combine data from two blocks for mean and variance.
See references to older papers/other work in the report:
- T. F. Chan, G. H. Golub, and R. J. LeVeque. Updating formulae and a pairwise algo- rithm for computing sample variances. Technical Report STAN-CS-79-773, Stanford University, Department of Computer Science, 1979.
- T. B. Terriberry. Computing higher-order moments online, 2008. http://people.xiph.org/~tterribe/notes/homs.html
- Philippe P. Pébay, Timothy B. Terriberry, Hemanth Kolla, and Janine C. Bennett: "Numerically Stable, Scalable Formulas for Parallel and Online Computation of Higher-Order Multivariate Central Moments with Arbitrary Weights." Computational Statistics, 31(4):1305–1325, December 2016. (PDF)
- West, D. H. D. "Updating mean and variance estimates: An improved method." Communications of the ACM 22.9 (1979): 532-535. https://dl.acm.org/citation.cfm?id=359153 (possibly of interest?)
- Chan, Tony F., Gene H. Golub, and Randall J. LeVeque. "Algorithms for computing the sample variance: Analysis and recommendations." The American Statistician 37.3 (1983): 242-247. https://amstat.tandfonline.com/doi/abs/10.1080/00031305.1983.10483115 (maybe not parallel but perhaps useful)
Expected behaviour
I can use a parallel version of MDAnalysis.analysis.rms.RMSF
Actual behaviour
No present.
Currently version of PMDA:
0.2.1