2006 Volume 35 Issue 2 Pages 79-91
Suppose that causal relationships among variables can be described by a causal diagram and the corresponding linear structural equation models. Intervention not to variables but to arrows will be considered, and the intervention effect can be formulated through the conditional intervention proposed by Pearl and Robins(1995). The identifiability criterion for the intervention to arrows is also investigated.The intervention effect on the variance of a response variable can be evaluated under the linear structural equation model.The intervention to arrows method may be useful to reduce variances of intermediate variables or covariates that cannot be controlled directly by an external action. A numerical example illustrates how to utilize the intervention to arrows method to reduce the variability of aresponse variable of interest.