Ancestor regression in linear structural equation models

Summary We present a new method for causal discovery in linear structural equation models. propose simple technique based on statistical testing models that can distinguish between ancestors and non-ancestors of any given variable. Naturally, this approach then be extended to estimating the order among all variables. provide explicit error control false discovery, at least asymptotically. This holds true even under Gaussianity, where other methods fail due non-identifiable structures. These Type I guarantees come cost reduced power. Additionally, we an asymptotically valid goodness-of-fit p-value assessing whether multivariate data stem from model.