Construction of a New Algebraic Statistical Theory for StatisticalCausal Inference

点击翻译按钮获取中文摘要

Conditional independence is of fundamental importance in statistical sciences. All statistical models, such as time series models or Bayesian models, are based on certain assumptions on conditional independence. Statistical conditional independence also characterizes the causal relations of a set of random variables. The main purpose of this research is to study the conditional independence using an axiomatic approach. We constructed a new statistical algebraic system called cain for studying statistical causal inference. This system is based on the basic properties of probability density functions which are essential for arguing for conditional independence. We further established an algebraic algorithm for deriving a set of relations concerning some conditional independence from other set of such conditions.