On the classes of causal networks, identifiable by simple independence tests

• Дата: 2018
• Автор: Balabanov, O.S.
• Формат: Стаття
• Мова: Ukrainian
• Опубліковано: Інститут програмних систем НАН України 2018
• Теми: causal network; 1-identifiable model; unconditional independence; conditional independence of 1st rank; locally minimal separator; collider; cycle; chain; d-separation; poly-tree; edge triangle; UDC 004.855:519.216
• Онлайн доступ: https://pp.isofts.kiev.ua/index.php/ojs1/article/view/281
• Назва журналу: Problems in programming

We tackle some theoretical problems of constraint-based approach to causal network inference from data (without prior restrictions). Our interest is to recover a model structure from independence tests of zero and first rank only. Class of 1-identifiable causal structures is defined. An idea to recognize whether model recovery is successfully completed (i.e. adequate model structure is outputted) is suggested. Theframework of locally minimal separation in DAG is shown to be appropriate instrument to tackle the problem. A few subclasses of class of 1-identifiable structures are specified; corresponding structural restrictions and criteria of recovery completeness are given. We present some causal structures which are not 1-identifiable.Problems in programming 2018; 2-3: 180-188