stablespec (version 0.1.3)
Stable Specification Search in Structural Equation Models
Description
An exploratory and heuristic approach for specification search
in Structural Equation Modeling. The basic idea is to subsample
the original data, and for each subset we search the
optimal models. Here the criteria of an optimal model is defined
by two objectives: to fit the data well and is simple (parsimonious) model.
As these objectives are conflicting, we apply NSGA-II to optimize,
such that to get optimal models for the whole range of model complexities.
From these optimal models, we observe the model specifications (structures)
that are both stable and parsimonious, called relevant. At the end we infer
causal model from these relevant structures.