On the intrinsic relations of correlations in some systems of linear structural equations

1Balabanov, OS
1Institute of Software Systems of the NAS of Ukraine, Kyiv
Dopov. Nac. akad. nauk Ukr. 2016, 12:17-21
https://doi.org/10.15407/dopovidi2016.12.017
Section: Information Science and Cybernetics
Language: Ukrainian
Abstract: 

For a probabilistic linear model of cyclic structure with four variables, we prove two simple inequality-type constraints on the set of correlations. Each of the inequalities (comprising two and three correlations, respectively) facilitates the rejection of the basic model in favor of an alternative model, which differs in that it contains an additional "diagonal" connection.

Keywords: correlation, hidden variable, inequality constraint, linear dependences, Markov property, model structure, model verification, relationship structure
References: 
  1. Scheines R., Spirtes P., Glymour C., Meek C., Richardson T. Multivar. Behavioral Res., 1998, 33, No 1: 65-118. doi: https://doi.org/10.1207/s15327906mbr3301_3, PMid:26771754
  2. Andon P. I., Balabanov O. S. Reports of the National Academy of Sciences of Ukraine, 2008, 9: 37-43 (in Ukrainian).
  3. Bell J. S. Physics, 1964, 1: 195-200.
  4. Tian J, Pearl J. Proc. 18th Conf. on Uncertainty in Artif. Intellig., 2002: 519-527.
  5. Kang Ch, Tian J. Proc. 22nd Conf. on Uncertainty in Artif. Intellig., 2006: 233-240.
  6. Ramsahai R. R. J. Mach. Learn. Res. — 2012. — 13: 829-848.
  7. Evans R. J. 22nd Workshop on Machine Learning and Signal Processing, 2012, 1-6 (Preprint: arXiv:1209.2978v1 [math.ST]).
  8. Drton M., Sturmfels B., Sullivant S. Probability Theory and Related Fields, 2007, 138, No 3-4: 463-493. doi: https://doi.org/10.1007/s00440-006-0033-2