DE: Generates a random model using RAM specification.
In dustinfife/DE: Performs an analysis of Data Elasticity

Description Usage Arguments Details Value Author(s) References Examples

This function generates a single random path model.

  DE(variable.names, paths, restrictions = NULL,
    prop.arrows = 0.2, allow.orphaned = FALSE,
    allow.bidir = FALSE, corr.exogenous = FALSE,
    corr.residuals = 0)

`variable.names`	A vector of variable names.
`paths`	The number of paths to be randomly generated. Can be a single value or a vector of two integers specifying a range.
`restrictions`	What kind of restrictions are set. (See details).
`prop.arrows`	What proportion of exogenous variables should be correlated. Defaults to .2.
`allow.orphaned`	Should orphaned variables be allowed when random models are generated?
`allow.bidir`	Should bidirectional arrows be allowed? (Note: this is not the same as a correlation).
`corr.exogenous`	Should exogenous variables be correlated?
`corr.residuals`	a value between 0 and 1 indicating the proportion of residuals the user allows to be correlated

Making restrictions is simple. For example, suppose one variable is Age. Obviously, Age should probably not be endogenous, so the user can specify Age as an endogenous variable. That is done by creating a matrix where the columns correspond to "From", "To", and "Include." For example, to specify that A must cause B, one would insert in the first row of the matrix c("A", "B", "1"). To specify that nothing can cause a variable (i.e., to make a variable exogenous), one would leave the "From" column as "". For example, the Age example would have c("", "Age", "0").

Allowing any variable to correlate with an endogenous variable is equivalent to correlating with the residuals of that endogenous variable. When the user specifies a non-zero value (k) for corr.residuals, the algorithm randomly selects k*(number of paths) of the paths to be double-headed, thereby permitting correlated residuals.

Returns a RAM matrix.

Dustin Fife

Fife, D.A., Rodgers, J.L., & Mendoza, J. L. (2013). Model conditioned data elasticity in path analysis: Assessing the "confoundability"

1
2
3

rest = matrix(c("A", "B", "1", 1,
						"", "A", "0", 1), nrow=2, byrow=TRUE)
DE(variable.names=LETTERS[1:6], paths=c(6,7), restrictions = NULL, prop.arrows = 0.2, allow.orphaned=FALSE, allow.bidir=FALSE, allow.cov.endogenous=FALSE)