pcalg (version 0.1-8)

beta.special.pcObj: Compute set of intervention effects in a fast way

Description

This function computes a set of intervention effects of one variable onto another variable in a fast way. The pc-object has to be precomputed.

Usage

beta.special.pcObj(x.pos,y.pos,pcObj,mcov=NA,amat=NA,amatSkel=NA,t.amat=NA)

Arguments

x.pos

Column of x in dat

y.pos

Column of y in dat

pcObj

Precomputed pc-object

mcov

covariance that was used in the pc-object fit

amat, amatSkel, t.amat

matrices that can be precomputed, if needed (see code for details on how to precompute)

Value

• estimates of intervention effects

Details

Estimate the intervention effect of x on y; the pcObj has to be precomputed. This method is intended to be a fast version of

beta.special(dat=NA,x.pos,y.pos,verbose=0,a=NA,myDAG=NA,myplot=FALSE, perfect=FALSE,method="local",collTest=TRUE,pcObj=pcObj,all.dags=NA, scaled.data=FALSE,u2pd="relaxed")

Thus, this is a faster version for the local method given a precomputed PC-Algo Object (relaxed udag2pdag, so CPDAG might not be a real CPDAG; this does not matter, since we try not to extend).

If this function is used repeatedly on the same pc-object (e.g. for different values of x and y), the matrices amat, amatSkel and t.amat can be precomputed in order to be more efficient (see the first entry in the code for furhter details).

References

M.H. Maathuis, M. Kalisch, P. B"uhlmann (2009), Estimating high-dimensional intervention effects from observational data; Annals of Statistics, 2009.

See Also

pcAlgo,dag2cpdag,beta.special.pcObj

Examples

set.seed(125)
p <- 10
n <- 10000

myDAG <- randomDAG(p, prob = 0.4)

cov.t <- trueCov(myDAG)
pcObj <- pcAlgo.Perfect(cov2cor(cov.t),directed=TRUE)
d.mat <- rmvDAG(n, myDAG, errDist = "normal")
cov.e <- cov(d.mat)
pcObj.est <- pcAlgo(d.mat,alpha=0.01,directed=TRUE)

x.pos <- 1
y.pos <- 5
true.eff <- causalEffect(myDAG,y.pos,x.pos)
est.eff.p <- beta.special.pcObj(x.pos=x.pos,y.pos=y.pos,pcObj=pcObj,mcov=cov.t)
est.eff.e <- beta.special.pcObj(x.pos=x.pos,y.pos=y.pos,pcObj=pcObj.est,mcov=cov.e)

true.eff
est.eff.p
est.eff.e