pwSEM (version 1.0.0)

CI.algorithm: The CI.algorithm function

Description

This function impliments the exploratory method of causal discovery called the CI (Causal Inference) algorithm of Pearl.

Usage

CI.algorithm(
  dat,
  family = NA,
  nesting = NA,
  smooth = TRUE,
  alpha.reject = 0.05,
  constrained.edges = NA,
  write.result = TRUE
)

Value

Just the partially-oriented graph output to the screen or just the adjacency matrix as a binary (0/1) matrix

Arguments

dat

A data frame containing the variables for which a partially- oriented dependency graph is sought. All variables in this data frame will be used except for those listed in the nesting= argument; these variables describe the nesting structure (if present) of the other variables.

family

A data frame giving the name of the distributional type of each variable that is not Gaussian. Example: data.frame(X1="binomial") if all variables except X1 are Gaussian but X1 is binomial.

nesting

A named list. Each name in the list is the name of the variable in the data set (except for the nesting= variables) followed by a character vector giving the names of the variables in dat holding the nesting structure. The default is nesting=NA, which is used if there are no nesting variables.

smooth

A logical value stating if linear (or generalized linear) links between the variables are assumed (smooth=FALSE) or not (smooth=FALSE). The default is smooth=TRUE, in which case generalized additive models are used to measure conditional independence and therefore non-linear relationships are assumed.

alpha.reject

A numerical value between 0 and 1 giving the "significance level" to use when judging (conditional) independence. The default value is alpha.reject=0.05.

constrained.edges

A character object giving the edges that must be constrained to not exist ("X|Y"), or X is a causal parent of Y ("X->Y") or X and Y have a common latent parent ("X<->Y"). Default is constrained.edges=NA (no edges are constrained).

write.result

A logical value indicating if you want the resulting partially-oriented dependency graph to be output to the screen (write.result = T) or just the adjacency matrix returned as output. The default value is write.result = T.

Examples

set.seed(11)
X1<-rnorm(500)
X2<-0.5*X1+rnorm(500,0,sqrt(1-0.5^2))
X3<-0.5*X2+rnorm(500,0,sqrt(1-0.5^2))
X4<-0.5*X2+rnorm(500,0,sqrt(1-0.5^2))
X5<-0.5*X3+0.5*X4+rnorm(500,0,sqrt(1-2*0.5^2))
my.dat<-data.frame(X1,X2,X3,X4,X5)
CI.algorithm(dat=my.dat,smooth=FALSE,alpha.reject=0.05)