I am using the grf R package (https://grf-labs.github.io/grf/reference/causal_forest.html) to obtain causal effect estimates for a continuous treatment variable. The package description says: "When W is continuous, we effectively estimate an average partial effect Cov[Y, W | X = x] / Var[W | X = x], and interpret it as a treatment effect given unconfoundedness." Does this mean that the average partial effect tau(X) can be interpreted as a "derivative" dY/dW or as another type of linear approximation to the causal effect "slope"? How can the absolute causal effect of the treatment be obtained from tau(X) (e.g. by integration)?